Encoding and decoding methods, and electronic devices

Abstract

Embodiments of this application provide an encoding and decoding method and an electronic device. The encoding method comprises the steps of: acquiring an image to be encoded; generating a feature map of C channels based on the image to be encoded, wherein the feature map includes feature values ​​of a plurality of feature points; generating an estimated information matrix of C channels based on the feature map of C channels; grouping the C channels into N channel groups; determining, for at least one target channel group among the N channel groups, at least one probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group based on at least one feature value of an encoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group; determining a probability distribution corresponding to an encoding target feature point based on at least one probability distribution parameter corresponding to an encoding target feature point; and encoding the encoding target feature point based on the probability distribution corresponding to the encoding target feature point. Thus, the at least one probability distribution parameter is determined based on information regarding a portion of the channels, thereby enabling improved encoding efficiency.

Description

Technology Field

[0001] This application claims priority to Chinese patent application No. 202210796212.X, filed with the Chinese Intellectual Property Office on July 7, 2022, under the title “Encoding and Decoding Method and Electronic Device,” which is incorporated herein by reference in its entirety.

[0002] The embodiments of this application relate to the field of encoding and decoding, in particular, encoding and decoding methods and electronic devices. Background Technology

[0003] The AI ​​(Artificial Intelligence) image compression algorithm is implemented based on deep learning and has a better compression effect than existing image compression technologies (e.g., JPEG (Joint Photographic Experts Group) and BPG (Better Portable Graphics). The process of compressing an image according to the AI ​​image compression algorithm predicts at least one probability distribution parameter corresponding to an encoding target point / decoding target point, then determines a probability distribution based on at least one probability distribution parameter, and then performs entropy encoding on the encoding target point / decoding target point based on the probability distribution to obtain a bitstream.

[0004] In conventional techniques, feature values ​​of encoded and decoded points across all channels are typically fused first to calculate context features of the target points for encoding and decoding. Subsequently, multi-layer convolution is performed on the context features and hyperprior features of the target points to determine at least one probability distribution parameter corresponding to the target points. However, the fusion of data from all channels and multiple convolutions requires a large amount of computation and takes a long time, which affects encoding and decoding efficiency.

[0005] This application provides an encoding and decoding method and an electronic device to solve the aforementioned technical problems. The encoding and decoding method can improve encoding and decoding efficiency.

[0006] According to the first aspect, an embodiment of the present application provides an encoding method. The method first acquires an image to be encoded; then, generates a feature map of C channels based on the image to be encoded, wherein the feature map includes feature values ​​of a plurality of feature points, and C is a positive integer; generates an estimated information matrix of C channels based on the feature map of C channels; then, groups the C channels into N channel groups, wherein N is an integer greater than 1, each channel group includes k channels, and the number of channels k included in any two channel groups is the same or different, and k is a positive integer; then, for at least one target channel group among the N channel groups, determines at least one probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group based on at least one feature value of at least one encoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group. Next, the method includes the step of determining a probability distribution corresponding to a feature point to be encoded based on at least one probability distribution parameter corresponding to a feature point to be encoded; and next, the step of encoding a feature point to be encoded based on the probability distribution corresponding to a feature point to be encoded.

[0007] Compared to existing techniques in which at least one probability distribution parameter is determined based on the feature value of an encoded feature point of all channels and the estimated information matrix of all channels, in this application, at least one probability distribution parameter corresponding to a feature point to be encoded needs to be determined solely based on at least one feature value of an encoded feature point within a feature map of a part of the channel and the estimated information matrix of the part of the channel. This can reduce the computing power required for encoding and improve encoding efficiency.

[0008] Additionally, compared to existing techniques where context features need to be generated based on at least one feature value of at least one encoded feature point, and then at least one probability distribution parameter is determined based on the context features and the estimated information matrix, in this application, context features do not need to be generated. This further reduces the computational power required for encoding and improves encoding efficiency.

[0009] Additionally, because the correlation between the feature maps of the channels is low, a larger amount of information is stored during compression. Therefore, in this application, the introduction of invalid information can be reduced and encoding performance can be improved.

[0010] For example, the feature map of each channel is And, where, "H" represents the height of the feature map of each channel and "W" represents the width of the feature map of each channel. The feature map of each channel may contain feature values ​​of H*W feature points.

[0011] For example, the estimated information matrix of each channel is It can include estimated information of H*W feature points.

[0012] For example, the estimated information may be information used to estimate probability distribution parameters and may include features and / or probability distribution parameters. This is not limited to this application.

[0013] For example, k = 1.

[0014] According to the first aspect, the step of determining at least one probability distribution parameter corresponding to an encoded feature point corresponding to a target channel group based on at least one feature value of at least one encoded feature point corresponding to a target channel group and an estimated information matrix corresponding to a target channel group comprises the step of performing linear weighting on at least one feature value of at least one encoded feature point corresponding to a target channel group and an estimated information matrix corresponding to a target channel group in order to determine at least one probability distribution parameter corresponding to an encoded feature point corresponding to a target channel group. Therefore, the amount of computation for determining at least one probability distribution parameter can be reduced from tens of thousands of original multiply-accumulate calculations to at least a few multiply-accumulate calculations and up to hundreds of multiply-accumulate calculations. This significantly reduces the amount of computation.

[0015] According to any one of the first aspect or the above-described embodiment of the first aspect, the estimated information matrix comprises estimated information of a plurality of feature points. To determine at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group, the step of performing linear weighting on at least one feature value of at least one encoded feature point corresponding to a target channel group and on the estimated information matrix corresponding to a target channel group comprises: a step of determining, based on the encoding target feature point, a first target region within a feature map corresponding to a target channel group and a second target region within the estimated information matrix corresponding to a target channel group; and a step of performing linear weighting on at least one feature value of at least one encoded feature point within the first target region and on the estimated information of at least one feature point within the second target region to obtain at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group. Therefore, the linear weighting is performed solely based on the feature value of a portion of the encoded feature points within the feature map and the estimated information of the feature point at a portion of the location within the estimated information matrix. This can reduce the computational power required to determine at least one probability distribution parameter and improve encoding efficiency.

[0016] For example, at least one probability distribution parameter may include at least one first probability distribution parameter and / or at least one second probability distribution parameter.

[0017] For example, the first probability distribution parameter is the mean, and the second probability distribution parameter is the variance.

[0018] According to any one of the first aspect or the above-described embodiments of the first aspect, when k exceeds 1, the step of determining a first target region in a feature map corresponding to a target channel group and a second target region in an estimated information matrix corresponding to a target channel group based on the encoding target feature point comprises: determining as the first target region an area of ​​a preset size centered on the encoding target feature point in the feature map of the first channel and an area of ​​a preset size centered on a feature point corresponding to the location of the encoding target feature point in the feature map of the second channel - the first channel corresponds to the encoding target feature point, and the second channel is a channel other than the first channel in the target channel group -; and determining as the second target region an area of ​​a preset size centered on the encoding target location in the estimated information matrix of the first channel and an area of ​​a preset size centered on the location corresponding to the encoding target location in the estimated information matrix of the second channel - the encoding target location is the location of the encoding target feature point -. Therefore, the probability distribution parameter is calculated based on at least one feature value of at least one encoded feature point surrounding the encoding target feature point, and the estimated information of the surrounding feature points, so that the calculated probability distribution parameter can be more accurate and the encoding quality can be further improved.

[0019] For example, the preset size may be the size of a linear weighting window, and may be ks1*ks2, where ks1 and ks2 are integers greater than 1, and ks1 and ks2 may be equal or not equal, and may be specifically set based on requirements. This is not limited in this application.

[0020] According to any one of the first aspect or the above-described embodiments of the first aspect, the step of performing linear weighting on at least one feature value of at least one encoded feature point within a first target region and estimated information of at least one feature point within a second target region to obtain at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group comprises: a step of determining a first target location based on a location other than an encoded location within the second target region—wherein the encoded location is at least one location of at least one encoded feature point—; and a step of performing linear weighting on at least one feature value of at least one encoded feature point within the first target region and estimated information of a feature point corresponding to the first target location to obtain at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group. Therefore, the linear weighting is performed solely based on the estimated information of a feature point at a part of a location within the second target region, thereby allowing the number of feature points for linear weighting to be reduced. This can reduce computational power and improve encoding efficiency.

[0021] According to any one of the first aspect or the above-described embodiment of the first aspect, when k exceeds 1, the step of determining a first target location based on a location other than an encoded location within a second target region comprises determining as the first target location a location corresponding to an encoded target location within a second target region within an estimated information matrix of a first channel and at least one other unencoded location, and a location corresponding to an encoded target location within a second target region within an estimated information matrix of a second channel and at least one other unencoded location, wherein the first channel corresponds to an encoded target feature point, and the second channel is a channel other than the first channel within a target channel group. When linear weighting is performed based only on a portion of other unencoded locations, the number of linear weighting feature points can be reduced. This can reduce computational power and improve encoding efficiency.

[0022] According to any one of the first aspect or the above-described embodiment of the first aspect, when k exceeds 1, the step of determining a first target location based on a location other than an encoded location within a second target region comprises determining as the first target location a location corresponding to an encoding target location within a second target region within an estimated information matrix of a first channel, and a location corresponding to an encoding target location within a second target region within an estimated information matrix of a second channel, wherein the first channel corresponds to an encoding target feature point, and the second channel is a channel other than the first channel within a target channel group. Therefore, the number of feature points for linear weighting can be reduced. This can further reduce the amount of computation for probability distribution parameters and improve encoding efficiency.

[0023] According to any one of the first aspect or the above-described embodiments of the first aspect, the step of performing linear weighting on at least one feature value of at least one encoded feature point within a first target region and estimated information of a feature point corresponding to a first target location in order to obtain at least one first probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group comprises: obtaining a preset weight matrix corresponding to a first channel corresponding to an encoding target feature point—wherein the preset weight matrix includes weight maps of k channels, and the size of the weight maps is the same as the size of the first target region—; and, based on the weight maps of k channels, performing linear weighting on at least one feature value of at least one encoded feature point within a first target region and estimated information of a feature point corresponding to a first target location in order to obtain at least one first probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group. Different channels correspond to the above-described preset weight matrix. Therefore, different probability distribution parameters can be obtained for encoding target feature points of different channels.

[0024] According to any one of the first aspect or the above-described embodiments of the first aspect, the estimated information matrix of C channels comprises a first feature matrix of C channels and a second feature matrix of C channels, wherein the first feature matrix comprises a first feature of a plurality of feature points and the second feature matrix comprises a second feature of a plurality of feature points, and at least one probability distribution parameter comprises at least one first probability distribution parameter and at least one second probability distribution parameter. Based on at least one feature value of at least one encoded feature point corresponding to a target channel group and the estimated information matrix corresponding to a target channel group, the step of determining at least one probability distribution parameter corresponding to an encoded feature point corresponding to a target channel group based on at least one feature value of at least one encoded feature point corresponding to a target channel group and the first feature matrix corresponding to a target channel group comprises: a step of determining at least one first probability distribution parameter corresponding to an encoded feature point corresponding to a target channel group based on at least one feature value of at least one encoded feature point corresponding to a target channel group and the first feature matrix corresponding to a target channel group; The method includes the step of determining at least one second probability distribution parameter corresponding to an encoded target feature point corresponding to a target channel group based on at least one feature value of at least one encoded feature point corresponding to a target channel group and a second feature matrix corresponding to a target channel group. A first feature matrix of C channels may be used to determine the first probability distribution parameter, and a second feature matrix of C channels may be used to determine the second probability distribution parameter. Therefore, both the first probability distribution parameter and the second probability distribution parameter can be corrected so that the obtained probability distribution parameter becomes more accurate. This improves the accuracy of the determined probability distribution.

[0025] According to any one of the first aspect or the above-described embodiments of the first aspect, the estimated information matrix of C channels comprises a first feature matrix of C channels and a second probability distribution parameter matrix of C channels, wherein the first feature matrix comprises a first feature of a plurality of feature points, the second probability distribution parameter matrix comprises a second probability distribution parameter of a plurality of feature points, and at least one probability distribution parameter comprises at least one first probability distribution parameter. The step of determining at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group based on at least one feature value of at least one encoded feature point corresponding to a target channel group and the estimated information matrix corresponding to the target channel group comprises the step of determining at least one first probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group based on at least one feature value of at least one encoded feature point corresponding to a target channel group and the first feature matrix corresponding to the target channel group. The step of determining a probability distribution corresponding to a feature point to be encoded based on at least one probability distribution parameter corresponding to a feature point to be encoded includes: determining at least one second probability distribution parameter corresponding to a feature point to be encoded corresponding to a target channel group based on a second probability distribution parameter matrix corresponding to a target channel group; and determining a probability distribution corresponding to a feature point to be encoded based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to a feature point to be encoded. Therefore, the first probability distribution parameter can be corrected so that the obtained probability distribution parameter becomes more accurate. This improves the accuracy of the determined probability distribution.

[0026] According to any one of the first aspect or the above-described embodiments of the first aspect, the estimated information matrix of C channels comprises a first probability distribution parameter matrix of C channels and a second feature matrix of C channels, wherein the first probability distribution parameter matrix comprises a first probability distribution parameter of a plurality of feature points, the second feature matrix comprises a second feature of a plurality of feature points, and at least one probability distribution parameter comprises at least one second probability distribution parameter. Based on at least one feature value of at least one encoded feature point corresponding to a target channel group and the estimated information matrix corresponding to the target channel group, the step of determining at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group comprises, based on at least one feature value of at least one encoded feature point corresponding to a target channel group and the second feature matrix corresponding to the target channel group, the step of determining at least one second probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group. The step of determining a probability distribution corresponding to a feature point to be encoded based on at least one probability distribution parameter corresponding to a feature point to be encoded includes: determining at least one first probability distribution parameter corresponding to a feature point to be encoded corresponding to a target channel group based on at least one first probability distribution parameter matrix corresponding to a target channel group; and determining a probability distribution corresponding to a feature point to be encoded based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to a feature point to be encoded. Therefore, the second probability distribution parameter can be corrected so that the obtained second probability distribution parameter becomes more accurate. This improves the accuracy of the determined probability distribution.

[0027] It should be understood that another estimated information matrix of C channels may be additionally generated based on the feature map of C channels, and then, at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group is determined based on at least one feature value of at least one encoded feature point corresponding to a target channel group, the estimated information matrix corresponding to the target channel group, and another estimated information matrix corresponding to the target channel group.

[0028] According to any one of the first aspect or the above-described embodiments of the first aspect, the step of determining at least one second probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group based on at least one feature value of at least one encoded feature point corresponding to a target channel group and a second feature matrix corresponding to a target channel group comprises: determining a first target region within a feature map corresponding to a target channel group and a third target region within a second feature matrix corresponding to a target channel group based on the encoding target feature point; determining at least one difference corresponding to at least one encoded feature point within the first target region based on at least one feature value of at least one encoded feature point within the first target region and at least one corresponding first probability distribution parameter; and performing linear weighting on a second feature of at least one feature point within the third target region and at least one difference corresponding to at least one encoded feature point within the first target region in order to obtain at least one second probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group. Therefore, the second probability distribution parameter can be calculated accurately.

[0029] For example, the first probability distribution parameter is the mean, and the second probability distribution parameter is the variance.

[0030] For example, at least one first probability distribution parameter corresponding to at least one encoded feature point can be obtained by first performing linear weighting on at least one feature value of at least one encoded feature point within a first target region and a first feature of a feature point corresponding to a first target location.

[0031] For example, when the estimated information matrix includes a first probability distribution parameter matrix and a second feature matrix, at least one second probability distribution parameter may also be determined in this manner. At least one first probability distribution parameter corresponding to an encoded feature point may be determined based on the first probability distribution parameter matrix.

[0032] For example, a method for determining at least one difference corresponding to at least one encoded feature point in a first target region based on at least one feature value of at least one encoded feature point in a first target region and at least one corresponding first probability distribution parameter may be to determine at least one difference between at least one feature value of at least one encoded feature point in the first target region and at least one corresponding first probability distribution parameter as at least one difference corresponding to at least one encoded feature point in the first target region.

[0033] For example, a method for determining at least one difference corresponding to at least one encoded feature point in a first target region based on at least one feature value of at least one encoded feature point in a first target region and at least one corresponding first probability distribution parameter may be to determine at least one absolute value of at least one difference between at least one feature value of at least one encoded feature point in the first target region and at least one corresponding first probability distribution parameter as at least one difference corresponding to at least one encoded feature point in the first target region.

[0034] For example, a method for determining at least one difference corresponding to at least one encoded feature point in a first target region based on at least one feature value of at least one encoded feature point in a first target region and at least one corresponding first probability distribution parameter may be to determine the square of at least one difference between at least one feature value of at least one encoded feature point in the first target region and at least one corresponding first probability distribution parameter as at least one difference corresponding to at least one encoded feature point in the first target region.

[0035] According to a second aspect, an embodiment of the present application provides a decoding method. The method first comprises the step of receiving a bitstream; then, the step of decoding the bitstream to obtain an estimated information matrix of C channels, where C is a positive integer; then, the step of decoding the bitstream to obtain a feature value of a feature point of C channels based on the estimated information matrix of C channels to obtain a feature map of C channels; for a feature point to be decoded, the step of determining a target channel group to which the channel corresponding to the feature point to be decoded belongs from N channel groups obtained by grouping C channels, wherein each channel group includes k channels, and the number of channels k included in any two channel groups is the same or different, k is a positive integer, and N is an integer greater than 1; and the step of determining at least one probability distribution parameter corresponding to the feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group. The method includes: a step of determining a probability distribution corresponding to a feature point to be decoded based on at least one probability distribution parameter corresponding to a feature point to be decoded; a step of decoding a feature point to be decoded based on the probability distribution corresponding to the feature point to be decoded in order to obtain a feature value; and a step of performing reconstruction based on a feature map of C channels in order to output a reconstructed image.

[0036] Compared to existing techniques in which probability distribution parameters are determined based on feature values ​​of decoded feature points of all channels and estimated information matrices of all channels, in this application, at least one probability distribution parameter corresponding to a feature point to be decoded needs to be determined solely based on at least one feature value of at least one decoded feature point within a feature map of a part of the channel and an estimated information matrix of the part of the channel. This can reduce the computational power required for decoding and improve decoding efficiency.

[0037] Additionally, compared to existing techniques where context features need to be generated based on at least one feature value of at least one decoded feature point, and then probability distribution parameters are determined based on the context features and the estimated information matrix, in this application, context features do not need to be generated. This further reduces the computational power required for decoding and improves decoding efficiency.

[0038] According to a second aspect, the step of determining at least one probability distribution parameter corresponding to a feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to a target channel group and an estimated information matrix corresponding to a target channel group comprises the step of performing linear weighting on at least one feature value of at least one decoded feature point corresponding to a target channel group and an estimated information matrix corresponding to a target channel group in order to determine at least one probability distribution parameter corresponding to a feature point to be decoded.

[0039] According to any one of the second aspect or the above-described embodiment of the second aspect, the estimated information matrix includes estimated information of a plurality of feature points. To determine at least one probability distribution parameter corresponding to a feature point to be decoded, the step of performing linear weighting on at least one feature value of at least one decoded feature point corresponding to a target channel group and on the estimated information matrix corresponding to the target channel group comprises: determining, based on the feature point to be decoded, a first target region within a feature map corresponding to a target channel group and a second target region within the estimated information matrix corresponding to a target channel group; and to obtain at least one probability distribution parameter corresponding to a feature point to be decoded, the step of performing linear weighting on at least one feature value of at least one decoded feature point within the first target region and on the estimated information of at least one feature point within the second target region.

[0040] According to any one of the above-described embodiments of the second aspect, when k exceeds 1, the step of determining a first target region in a feature map corresponding to a target channel group and a second target region in an estimated information matrix corresponding to a target channel group based on a decoding target feature point is:

[0041] The method includes the step of determining as a first target area an area of ​​a preset size centered on a feature point to be decoded within a feature map of a first channel, and an area of ​​a preset size centered on a feature point corresponding to the location of the feature point to be decoded within a feature map of a second channel - wherein the first channel corresponds to the feature point to be decoded, and the second channel is a channel other than the first channel within a target channel group -; and the step of determining as a second target area an area of ​​a preset size centered on a location to be decoded within an estimated information matrix of a first channel, and an area of ​​a preset size centered on a location corresponding to the location to be decoded within an estimated information matrix of a second channel - wherein the location to be decoded is the location of the feature point to be decoded -

[0042] According to any one of the embodiments of the second aspect or the above-described embodiment of the second aspect, the step of performing linear weighting on at least one feature value of at least one decoded feature point in a first target region and the estimated information of at least one feature point in a second target region to obtain at least one probability distribution parameter corresponding to a feature point to be decoded comprises: a step of determining a first target location based on a location other than a decoded location in the second target region, wherein the decoded location is at least one location of at least one decoded feature point; and a step of performing linear weighting on at least one feature value of at least one decoded feature point in a first target region and the estimated information of a feature point corresponding to the first target location to obtain at least one probability distribution parameter corresponding to a feature point to be decoded.

[0043] According to any one of the embodiments of the second aspect or the above-described embodiment of the second aspect, when k exceeds 1, the step of determining a first target location based on a location other than a decoded location within a second target region comprises determining as the first target location a location corresponding to a decoding target location within a second target region within an estimated information matrix of a first channel and at least one other undecoded location, and a location corresponding to a decoding target location within a second target region within an estimated information matrix of a second channel and at least one other undecoded location, wherein the first channel corresponds to a decoding target feature point and the second channel is a channel other than the first channel within a target channel group.

[0044] According to any one of the above-described embodiments of the second aspect or the second aspect, when k exceeds 1, the step of determining a first target location based on a location other than a decoded location within a second target region comprises determining a location corresponding to a decoding target location within a second target region within an estimated information matrix of a first channel, and a location corresponding to a decoding target location within a second target region within an estimated information matrix of a second channel as the first target location, wherein the first channel corresponds to a decoding target feature point, and the second channel is a channel other than the first channel within a target channel group.

[0045] According to any one of the embodiments of the second aspect or the above-described embodiment of the second aspect, the step of performing linear weighting on at least one feature value of at least one decoded feature point within a first target region and estimated information of a feature point corresponding to a first target location in order to obtain at least one probability distribution parameter corresponding to a feature point to be decoded comprises: the step of obtaining a preset weight matrix corresponding to a first channel corresponding to a feature point to be decoded—the preset weight matrix includes a weight map of k channels—; and the step of performing linear weighting on at least one feature value of at least one decoded feature point within a first target region and estimated information of a feature point corresponding to a first target location based on the weight map of k channels in order to obtain at least one probability distribution parameter corresponding to a feature point to be decoded.

[0046] According to any one of the embodiments of the second aspect or the above-described embodiment of the second aspect, the estimated information matrix of C channels comprises a first feature matrix of C channels and a second feature matrix of C channels, wherein the first feature matrix comprises a first feature of a plurality of feature points and the second feature matrix comprises a second feature of a plurality of feature points, and at least one probability distribution parameter comprises at least one first probability distribution parameter and at least one second probability distribution parameter. The step of determining at least one probability distribution parameter corresponding to a feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to a target channel group and the estimated information matrix corresponding to a target channel group comprises the step of determining at least one first probability distribution parameter corresponding to a feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to a target channel group and the first feature matrix corresponding to a target channel group; The method includes the step of determining at least one second probability distribution parameter corresponding to a decoding target feature point based on at least one feature value of at least one decoded feature point corresponding to a target channel group and a second feature matrix corresponding to a target channel group.

[0047] According to any one of the embodiments of the second aspect or the above-described embodiment of the second aspect, the estimated information matrix of C channels comprises a first feature matrix of C channels and a second probability distribution parameter matrix of C channels, wherein the first feature matrix comprises a first feature of a plurality of feature points, the second probability distribution parameter matrix comprises a second probability distribution parameter of a plurality of feature points, and at least one probability distribution parameter comprises at least one first probability distribution parameter. The step of determining at least one probability distribution parameter corresponding to a feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to a target channel group and the estimated information matrix corresponding to the target channel group comprises the step of determining at least one first probability distribution parameter corresponding to a feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to a target channel group and the first feature matrix corresponding to the target channel group. The step of determining a probability distribution corresponding to a feature point to be decoded based on at least one probability distribution parameter corresponding to a feature point to be decoded includes: determining at least one second probability distribution parameter of a feature point to be decoded based on at least one second probability distribution parameter matrix corresponding to a target channel group; and determining a probability distribution corresponding to a feature point to be decoded based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to a feature point to be decoded.

[0048] According to any one of the embodiments of the second aspect or the above-described embodiment of the second aspect, the estimated information matrix of C channels comprises a first probability distribution parameter matrix of C channels and a second feature matrix of C channels, wherein the first probability distribution parameter matrix comprises a first probability distribution parameter of a plurality of feature points, the second feature matrix comprises a second feature of a plurality of feature points, and at least one probability distribution parameter comprises at least one second probability distribution parameter. The step of determining at least one probability distribution parameter corresponding to a feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to a target channel group and the estimated information matrix corresponding to the target channel group comprises the step of determining at least one second probability distribution parameter corresponding to a feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to a target channel group and the second feature matrix corresponding to the target channel group. The step of determining a probability distribution corresponding to a feature point to be decoded based on at least one probability distribution parameter corresponding to a feature point to be decoded includes: determining at least one first probability distribution parameter corresponding to a feature point to be decoded based on a first probability distribution parameter matrix corresponding to a target channel group; and determining a probability distribution corresponding to a feature point to be decoded based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to a feature point to be decoded.

[0049] According to any one of the embodiments of the second aspect or the above-described embodiment of the second aspect, the step of determining at least one second probability distribution parameter corresponding to a feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to a target channel group and a second feature matrix corresponding to a target channel group comprises: determining a first target region within a feature map corresponding to a target channel group and a third target region within a second feature matrix corresponding to a target channel group based on the feature point to be decoded; determining at least one difference corresponding to at least one decoded feature point within the first target region based on at least one feature value of at least one decoded feature point within the first target region and at least one corresponding first probability distribution parameter; and performing linear weighting on the second feature of at least one feature point within the third target region and at least one difference corresponding to at least one decoded feature point within the first target region in order to obtain at least one second probability distribution parameter corresponding to a feature point to be decoded.

[0050] Any one of the second aspect and any one of the embodiment of the second aspect corresponds to any one of the first aspect and any one of the embodiment of the first aspect. For the technical effect corresponding to any one of the second aspect and any one of the embodiment of the second aspect, reference is made to the technical effect corresponding to any one of the first aspect and any one of the embodiment of the first aspect. Details are not described again in this specification.

[0051] According to a third aspect, an embodiment of this application provides an encoder configured to perform an encoding method in any one of the first aspect and an embodiment of the first aspect.

[0052] Any one of the third aspect and any one of the embodiment of the third aspect corresponds to any one of the first aspect and any one of the embodiment of the first aspect. For the technical effect corresponding to any one of the third aspect and any one of the embodiment of the third aspect, refer to the technical effect corresponding to any one of the first aspect and any one of the embodiment of the first aspect. Details are not described again in this specification.

[0053] According to the fourth aspect, an embodiment of this application provides a decoder configured to perform a decoding method in any one of the second aspect and an embodiment of the second aspect.

[0054] Any one of the fourth aspect and any one of the embodiment of the fourth aspect corresponds to any one of the second aspect and any one of the embodiment of the second aspect. For the technical effect corresponding to any one of the fourth aspect and any one of the embodiment of the fourth aspect, reference is made to the technical effect corresponding to any one of the second aspect and any one of the embodiment of the second aspect. Details are not described again in this specification.

[0055] According to the fifth aspect, an embodiment of this application provides an electronic device comprising a memory and a processor. The memory is coupled to the processor. The memory stores program instructions. When the program instructions are executed by the processor, the electronic device may perform an encoding method in the first aspect or any one of the possible embodiments of the first aspect.

[0056] Any one of the fifth aspect and any one of the embodiment of the fifth aspect corresponds to any one of the first aspect and any one of the embodiment of the first aspect. For the technical effect corresponding to any one of the fifth aspect and any one of the embodiment of the fifth aspect, refer to the technical effect corresponding to any one of the first aspect and any one of the embodiment of the first aspect. Details are not described again in this specification.

[0057] According to the sixth aspect, an embodiment of this application provides an electronic device comprising a memory and a processor. The memory is coupled to the processor. The memory stores program instructions. When the program instructions are executed by the processor, the electronic device may perform a decoding method in the second aspect or any one of the possible embodiments of the second aspect.

[0058] Any one of the sixth aspect and any one of the embodiment of the sixth aspect corresponds to any one of the second aspect and any one of the embodiment of the second aspect. For the technical effect corresponding to any one of the sixth aspect and any one of the embodiment of the sixth aspect, refer to the technical effect corresponding to any one of the second aspect and any one of the embodiment of the second aspect. Details are not described again in this specification.

[0059] According to the seventh aspect, an embodiment of this application provides a chip comprising one or more interface circuits and one or more processors. The interface circuit is configured to receive a signal from the memory of an electronic device and to transmit the signal to the processor. The signal includes computer instructions stored in the memory. When the processor executes program instructions, the electronic device may perform an encoding method in the first aspect or any one of the possible embodiments of the first aspect.

[0060] Any one of the seventh aspect and any one of the embodiment of the seventh aspect corresponds to any one of the first aspect and any one of the embodiment of the first aspect. For the technical effect corresponding to any one of the seventh aspect and any one of the embodiment of the seventh aspect, refer to the technical effect corresponding to any one of the first aspect and any one of the embodiment of the first aspect. Details are not described again in this specification.

[0061] According to the eighth aspect, an embodiment of this application provides a chip comprising one or more interface circuits and one or more processors. The interface circuit is configured to receive a signal from the memory of an electronic device and to transmit the signal to the processor. The signal includes computer instructions stored in the memory. When the processor executes program instructions, the electronic device may perform a decoding method in the second aspect or any one of the possible embodiments of the second aspect.

[0062] Any one of the eighth aspect and any one of the embodiment of the eighth aspect corresponds to any one of the second aspect and any one of the embodiment of the second aspect. For the technical effect corresponding to any one of the eighth aspect and any one of the embodiment of the eighth aspect, refer to the technical effect corresponding to any one of the second aspect and any one of the embodiment of the second aspect. Details are not described again in this specification.

[0063] According to the ninth aspect, an embodiment of this application provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is operated on a computer or processor, the computer or processor may perform an encoding method in any one of the first aspect or a possible embodiment of the first aspect.

[0064] Any one of the ninth aspect and any one of the embodiment of the ninth aspect corresponds to any one of the first aspect and any one of the embodiment of the first aspect. For the technical effect corresponding to any one of the ninth aspect and any one of the embodiment of the ninth aspect, refer to the technical effect corresponding to any one of the first aspect and any one of the embodiment of the first aspect. Details are not described again in this specification.

[0065] According to the tenth aspect, an embodiment of this application provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is operated on a computer or processor, the computer or processor may perform a decoding method in any one of the second aspect or a possible embodiment of the second aspect.

[0066] Any one of the tenth aspect and any one of the embodiment of the tenth aspect corresponds to any one of the second aspect and any one of the embodiment of the second aspect. For the technical effect corresponding to any one of the tenth aspect and any one of the embodiment of the tenth aspect, reference is made to the technical effect corresponding to any one of the second aspect and any one of the embodiment of the second aspect. Details are not described again in this specification.

[0067] According to the eleventh aspect, an embodiment of this application provides a computer program product. The computer program product includes a software program. When the software program is executed by a computer or a processor, the computer or the processor may perform an encoding method in any one of the first aspect or a possible embodiment of the first aspect.

[0068] Any one of the eleventh aspect and any one of the eleventh aspect corresponds to any one of the first aspect and any one of the first aspect. For the technical effect corresponding to any one of the eleventh aspect and any one of the eleventh aspect, refer to the technical effect corresponding to any one of the first aspect and any one of the first aspect. Details are not described again in this specification.

[0069] According to the 12th aspect, an embodiment of this application provides a computer program product. The computer program product includes a software program. When the software program is executed by a computer or a processor, the computer or the processor may perform a decoding method in any one of the second aspect or a possible embodiment of the second aspect.

[0070] Any one of the 12th aspect and any one of the 12th aspect corresponds to any one of the 2nd aspect and any one of the 2nd aspect. For the technical effect corresponding to any one of the 12th aspect and any one of the 12th aspect, refer to the technical effect corresponding to any one of the 2nd aspect and any one of the 2nd aspect. Details are not described again in this specification.

[0071] According to the 13th aspect, an embodiment of this application provides a bitstream generation method used to generate a bitstream according to an encoding method in any one of the first aspect and an embodiment of the first aspect.

[0072] According to the 14th aspect, an embodiment of this application provides a bitstream storage method used to store a bitstream generated according to a bitstream generation method in any one of the 13th aspect and an embodiment of the 13th aspect.

[0073] According to the 15th aspect, an embodiment of this application provides a bitstream transmission method used to transmit a bitstream generated according to a bitstream generation method in any one of the 13th aspect and an embodiment of the 13th aspect. Brief explanation of the drawing

[0074] Figure 1 is a diagram of an example of the structure of a system framework. Figure 2 is a diagram of an example of an encoding process. Figure 3a is a diagram of an example of a channel group. Figure 3b is a diagram of an example of a channel group. Figure 3c is a diagram of an example of the encoding order. Figures 4a and 4b are diagrams of an example of a decoding process. Figure 5 is a diagram of an example of an end-to-end image compression framework. Figure 6 is a diagram of an example of an encoding process. Figure 7 is a diagram of an example of a decoding process. Figure 8 is a diagram of an example of an encoding process. Figure 9 is a diagram of an example of a decoding process. Figure 10 is a diagram of an example of an encoding process. Figure 11 is a diagram of an example of a decoding process. Figure 12a is a diagram of an example of a process for determining probability distribution parameters. Figure 12b is a diagram of an example of a process for determining probability distribution parameters. Figure 12c is a diagram of an example of a process for determining probability distribution parameters. Figure 12d is a diagram of an example of a process for determining probability distribution parameters. Figure 13a is a diagram of an example of a process for determining probability distribution parameters. Figure 13b is a diagram of an example of a process for determining probability distribution parameters. Figure 14a is a diagram of an example of a process for determining probability distribution parameters. Figure 14b is a diagram of an example of a process for determining probability distribution parameters. Figure 15 is a diagram of an example of the structure of the device. Specific details for implementing the invention

[0075] The following clarifies the technical solution of the embodiments of this application with reference to the attached drawings in the embodiments of this application. It is clear that the described embodiments are only a part, not all, of the embodiments of this application. All other embodiments obtained by a person skilled in the art based on the embodiments of this application without creative effort will fall within the scope of protection of this application.

[0076] The term “and / or” in this specification describes only the association relationship for the associated objects and indicates that three relationships may exist. For example, A and / or B may indicate the following three cases: only A exists, both A and B exist, and only B exists.

[0077] In the specification and claims of the embodiments of this application, terms such as "first," "second," etc., are intended to distinguish between different objects but do not indicate a specific order of objects. For example, the first target object and the second target object are used to distinguish between different target objects but are not used to describe a specific order of target objects.

[0078] In the embodiments of this application, words such as "example," "for example," etc. are used to indicate providing an example, illustration, or explanation. Any embodiment or design method described as "example" or "for example" in the embodiments of this application should not be described as having more desirable or greater advantages than other embodiments or design methods. Precisely, the use of words such as "example," "for example," etc. is intended to present related concepts in a specific manner.

[0079] In the description of the embodiments of this application, unless otherwise stated, “a plurality of” means two or more than two. For example, a plurality of processing units is two or more processing units, and a plurality of systems is two or more systems.

[0080] FIG. 1 is a diagram of an example of the structure of a system framework. It should be understood that the system illustrated in FIG. 1 is merely an example, and that the system in this application may have more or fewer components than those illustrated in the drawings, may combine two or more components, or have different component configurations. The various components illustrated in FIG. 1 may be implemented in hardware, software, or a combination of hardware and software, including one or more signal processing and / or application-specific integrated circuits.

[0081] Refer to Fig. 1. For example, an image encoding process may be as follows: an image to be encoded is input into an AI encoding unit and processed by the AI ​​encoding unit, and the feature values ​​of the feature points to be encoded and their corresponding probability distributions are output. Next, the feature values ​​of the feature points to be encoded and their corresponding probability distributions are input into an entropy encoding unit. To output a bitstream, the entropy encoding unit performs entropy encoding on the feature values ​​of the feature points to be encoded based on the probability distributions corresponding to the feature points to be encoded.

[0082] Refer to FIG. 1. For example, an image decoding process may be as follows: After acquiring a bitstream, an entropy decoding unit may perform entropy decoding on a feature point to be decoded based on a probability distribution predicted by the AI ​​decoding unit based on at least one feature value of at least one decoded feature point corresponding to the feature point to be decoded, in order to output at least one feature value of at least one decoded feature point to the AI ​​decoding unit. After entropy decoding for all feature points to be decoded is completed, the AI ​​decoding unit performs reconstruction based on at least one feature value corresponding to at least one decoded feature point in order to output a reconstructed image.

[0083] For example, entropy encoding is an encoding without any loss of information based on the entropy principle in the encoding process. Entropy encoding may include multiple types, for example, Shannon encoding, Huffman encoding, and arithmetic encoding (arithmetic coding). This is not limited to this application.

[0084] For example, the encoding target image input to the AI ​​encoding unit may be any one of a raw (unprocessed) image, an RGB (Red Green Blue) image, or a YUV image ("Y" represents Luminance (or Luma), and "U" and "V" represent Chrominance and Chroma, respectively). This is not limited to this application.

[0085] For example, the compression process and the decompression process may be performed by the same electronic device or by different electronic devices. This is not limited to this application.

[0086] For example, this application may be applied to the compression and decompression of a single image or to the compression and decompression of multiple frames of an image within a video sequence. This is not limited to this application.

[0087] For example, this application may be applicable to multiple scenarios, e.g., Huawei image (or video) cloud storage (or transmission) scenarios, video surveillance scenarios, or live broadcast scenarios. This is not limited to this application.

[0088] Figure 2 is a diagram of an example of an encoding process.

[0089] S201: Acquired the image to be encoded.

[0090] For example, the encoder side can acquire an image to be encoded, and then, in order to acquire a corresponding bitstream, can encode the image to be encoded by referring to S202 to S206.

[0091] S202: Generate a feature map of C channels based on an image to be encoded, wherein the feature map includes feature values ​​of multiple feature points.

[0092] For example, to transform an image to be encoded into another space, a spatial transformation can be performed on the image to be encoded, thereby reducing the temporal and spatial redundancy of the image to be encoded, and obtaining a feature map of C (where C is a positive integer) channels.

[0093] For example, the feature map of each channel is "H" represents the height of the feature map of each channel, and "W" represents the width of the feature map of each channel. The feature map of each channel may contain feature values ​​of H*W feature points.

[0094] For example, FIG. 2 illustrates the feature map of channel 1, the feature map of channel 2, ..., the feature map of channel k, the feature map of channel c1, ..., the feature map of channel ck, ..., the feature map of channel C-1, and the feature map of channel C among C channel feature maps. Both c1 and ck are positive integers less than C.

[0095] S203: Generates an estimated information matrix of C channels based on feature maps of C channels.

[0096] For example, after C channel feature maps are acquired, feature extraction can be performed on the C channel feature maps to acquire the estimated information matrices of C channels. For example, the estimated information matrices of each channel are It can include estimated information of H*W feature points.

[0097] For example, the estimated information may be information used to estimate probability distribution parameters and may include features and / or probability distribution parameters. This is not limited to this application.

[0098] For example, FIG. 2 illustrates the estimated information matrix of channel 1, the estimated information matrix of channel 2, ..., the estimated information matrix of channel k, the estimated information matrix of channel c1, ..., the estimated information matrix of channel ck, ..., the estimated information matrix of channel C-1, and the estimated information matrix of channel C among the estimated information matrices of C channels.

[0099] S204: Groups C channels into N channel groups.

[0100] For example, C channels may be grouped into at least N channel groups (where N is an integer greater than 1), and each channel group may include at least one channel. The number of channels included in one channel group is represented by k (where k is a positive integer less than C), and it is assumed that k within all channel groups may be the same or different. This is not limited in this application.

[0101] FIG. 3a is a diagram of an example of a channel group. In the embodiment of FIG. 3a, the number of channels k included in all channel groups is the same.

[0102] Refer to Fig. 3a. For example, k = 2. Specifically, every 2 of C channels are used to form one channel group. Therefore, the number of acquired channel groups is N = C / 2. It is assumed that C = 192 and k = 2. In this case, N = 96. In other words, to acquire 96 channel groups, every 2 of 192 channels can be used to form one channel group.

[0103] FIG. 3b is a diagram of an example of a channel group. In the embodiment of FIG. 3b, the number of channels k included in all channel groups is shown as different.

[0104] Refer to FIG. 3b. For example, among C channels, Channel 1 can be used to form Channel Group 1, among C channels, Channels 2 and 3 can be used to form Channel Group 2, among C channels, Channels 4, 5, and 6 can be used to form Channel Group 3, ..., among C channels, Channels C-1 and C can be used to form Channel Group N.

[0105] FIGS. 3a and FIGS. 3b are merely examples of this application, and it should be noted that k can alternatively be set to other values ​​as required. This is not limited to this application.

[0106] It should be noted that different channel groups may include the same channel. For example, channel group 1 may include channel 1 and channel 2, and channel group 2 may include channel 2 and channel 3. This is not limited to this application.

[0107] S205: For at least one target channel group among N channel groups, at least one probability distribution parameter corresponding to an encoded target feature point corresponding to the target channel group is determined based on at least one feature value of at least one encoded feature point corresponding to the target channel group and an estimated information matrix corresponding to the target channel group.

[0108] For example, all of N channel groups may be sequentially determined as target channel groups, and then, for each target channel group, at least one probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group may be determined based on at least one feature value of at least one encoded feature point corresponding to the target channel group and an estimated information matrix corresponding to the target channel group. In this application, one target channel group among the N channel groups is used as an example to explain the process of determining a probability distribution parameter corresponding to one encoding target feature point within a feature map of channel c within the target channel group.

[0109] Figure 3c is a diagram of an example of the encoding order.

[0110] (1) in FIG. 3c illustrates a feature map of one channel. The size of the feature map is 10*10, and each block represents one feature point. For example, the encoder side can sequentially encode all feature points within the feature map in the order shown in (1) in FIG. 3c, specifically, it can sequentially encode all feature points from left to right starting from the first row until all feature points within the feature map are encoded, and after all feature points within the first row are encoded, it can start encoding all feature points within the second row from left to right, and so on.

[0111] (2) in FIG. 3c illustrates a feature map of one channel. The size of the feature map is 10*10, and each block represents one feature point. White blocks represent unencoded feature points, and black blocks represent encoded feature points (the encoder side first divides the feature map of each channel based on a black-and-white checkerboard, and the encoder side first encodes the feature points corresponding to the black blocks, and then encodes the feature points corresponding to the white blocks). In a possible embodiment, in the process of encoding the feature points corresponding to the black blocks, the corresponding probability distribution parameter is determined based on the estimated information matrix; in the process of encoding the feature points corresponding to the white blocks, the corresponding probability distribution parameter is determined based on S204 of this application. In a possible embodiment, in both the process of encoding the feature points corresponding to the black blocks and the process of encoding the feature points corresponding to the white blocks, the corresponding probability distribution parameter is determined based on S204 of this application.

[0112] For example, regarding feature points corresponding to white blocks, the encoder side can sequentially encode feature points corresponding to all white blocks in the feature map in the order of (2) in FIG. 3c, and specifically, until feature points corresponding to all white blocks in the feature map are encoded, feature points corresponding to all white blocks can be sequentially encoded from left to right starting from the first row, and after feature points corresponding to all white blocks in the first row are encoded, feature points corresponding to all white blocks in the second row can be encoded from left to right, and so on.

[0113] For example, in the process where the encoder side determines the corresponding probability distribution parameter based on S204 in this application in the process of encoding feature points corresponding to a black block, the order of encoding feature points corresponding to a black block is similar to the order of encoding feature points corresponding to a white block. Details are not described again in this specification.

[0114] (2) in FIG. 3c is merely an example, and it should be understood that the positions of the black blocks and white blocks within the black-and-white checkerboard can alternatively be interchanged. This is not limited to this application.

[0115] For example, the feature point to be encoded may be selected from the unencoded feature points within the feature map of channel c in the encoding order of FIG. 3c. Next, at least one encoded feature point within the feature map corresponding to the target channel group may be determined, and then, in order to determine at least one probability distribution parameter corresponding to one feature point to be encoded within the feature map of channel c within the target channel group, a weighting calculation is performed on at least one feature value of at least one encoded feature point within the feature map corresponding to the target channel group and on the estimated information matrix corresponding to the target channel group. The feature map corresponding to the target channel group may be a feature map of k channels included within the target channel group. The estimated information matrix corresponding to the target channel group may be an estimated information matrix of k channels included within the target channel group. Therefore, at least one probability distribution parameter corresponding to one feature point to be encoded within the feature map of channel c may be determined. Therefore, in order to determine at least one probability distribution parameter corresponding to each encoding target feature point within the feature map of channel c, the unencoded feature points within the feature map of channel c are sequentially determined as encoding target feature points.

[0116] Therefore, at least one probability distribution parameter corresponding to each encoding target feature point within the feature map of each channel within the target channel group can be determined.

[0117] In the manner described above, probability distribution parameters corresponding to encoding target feature points corresponding to all target channel groups, for example, at least one probability distribution parameter corresponding to an encoding target feature point corresponding to channel group 1, at least one probability distribution parameter corresponding to an encoding target feature point corresponding to channel group 2, ..., and at least one probability distribution parameter corresponding to an encoding target feature point corresponding to channel group N can be determined.

[0118] S206: Determines a probability distribution corresponding to a feature point to be encoded based on at least one probability distribution parameter corresponding to a feature point to be encoded.

[0119] For example, after at least one probability distribution parameter corresponding to an encoding target feature point is determined, for each encoding target feature point, a probability distribution corresponding to each encoding target feature point can be determined based on at least one probability distribution parameter corresponding to each encoding target feature point.

[0120] For example, at least one probability distribution parameter is a Gaussian distribution parameter, and the determined probability distribution corresponding to the feature points to be encoded is a Gaussian probability distribution.

[0121] S207: To obtain a bitstream, encoding target feature points based on a probability distribution corresponding to the encoding target feature points.

[0122] For example, for each feature point to be encoded, the feature value of each feature point to be encoded can be encoded based on a probability distribution corresponding to each feature point to be encoded, and the bitstream of the image to be encoded can be obtained after the feature values ​​of all feature points within a feature map of C channels have been encoded.

[0123] For example, the encoder side may store the bitstream of the image to be encoded locally or transmit the bitstream of the image to be encoded to the decoder side. This is not limited to this application.

[0124] For example, the encoder side may generate prior information of C channels based on feature maps of C channels, and then perform reconstruction based on the prior information of C channels to obtain an estimated information matrix of C channels. For example, the encoder side may alternatively encode the prior information of C channels to obtain a bitstream of the prior information of C channels; then, the bitstream of the prior information may be stored or transmitted to the decoder side, so that when the bitstream of the image to be encoded is sequentially decoded, the estimated information matrix may be determined based on the prior information, and then at least one probability distribution parameter corresponding to the feature point to be decoded is determined based on the estimated information matrix.

[0125] Compared to existing techniques in which probability distribution parameters are determined based on feature values ​​of encoded feature points of all channels and estimated information matrices of all channels, in this application, at least one probability distribution parameter corresponding to a feature point to be encoded needs to be determined solely based on at least one feature value of at least one encoded feature point within a feature map of a part of the channel and the estimated information matrix of the part of the channel. This can reduce the computational power required for encoding and improve encoding efficiency.

[0126] Additionally, compared to existing techniques where context features need to be generated based on at least one feature value of at least one encoded feature point, and then at least one probability distribution parameter is determined based on the context features and the estimated information matrix, in this application, context features do not need to be generated. This further reduces the computational power required for encoding and improves encoding efficiency.

[0127] Additionally, because the correlation between the feature maps of the channels is low, a larger amount of information is stored during compression. Therefore, in this application, the introduction of invalid information can be reduced and encoding performance can be improved.

[0128] Figures 4a and 4b are diagrams of an example of a decoding process.

[0129] S401: Received bitstream.

[0130] For example, the bitstream received by the decoder side may include the image bitstream and the bitstream corresponding to the image among the bitstreams of previous information of C channels.

[0131] For example, the image bitstream may contain encoded data of feature values ​​of feature points of C channels, and the bitstream of previous information of C channels may contain encoded data of previous information of feature points of C channels.

[0132] S402: Decode the bitstream to obtain the estimated information matrix of C channels.

[0133] For example, after a bitstream is received, the bitstream may be parsed to obtain encoded data of previous information of feature points of C channels, and then, the encoded data of previous information of feature points of C channels is processed to obtain an estimated information matrix of C channels.

[0134] For example, FIGS. 4a and 4b illustrate the estimated information matrix of channel 1, the estimated information matrix of channel 2, ..., the estimated information matrix of channel c1, the estimated information matrix of channel c2, ..., the estimated information matrix of channel ck, ..., the estimated information matrix of channel C-1, and the estimated information matrix of channel C among the estimated information matrices of C channels. Both c1 and ck are positive integers less than C.

[0135] S403: Decode the bitstream to obtain feature values ​​of feature points of C channels based on the estimated information matrix of C channels, in order to obtain feature maps of C channels.

[0136] For example, in a process of parsing a bitstream, the bitstream may be further parsed to obtain encoded data of feature values ​​of feature points of C channels, and then, in order to obtain feature values ​​of feature points of C channels, i.e., feature maps of C channels, the encoded data of feature values ​​of feature points of C channels may be decoded based on an estimated information matrix of C channels. In this application, an example in which one decoding target feature point of one channel is decoded is used for illustrative purposes.

[0137] For example, the decoding order on the decoder side is the same as the encoding order on the encoder side. For details, refer to the description in the embodiment of FIG. 3c. Details are not described again in this specification.

[0138] For example, the feature point to be decoded can be determined by selecting from the undecoded feature point corresponding to the channel in a decoding order corresponding to the encoding order in FIG. 3c. Next, the feature value of the feature point to be decoded can be obtained through decoding with reference to the following steps (S4031 to S4034).

[0139] S4031: From N channel groups obtained by grouping C channels, determine the target channel group to which the channel corresponding to the feature point to be decoded belongs.

[0140] For example, the decoder side may alternatively group C channels into N channel groups, each channel group may include k channels. Details are similar to the method described above for the encoder side to group channel groups. Details are not described again in this specification.

[0141] For example, a target channel group to which the channel corresponding to the feature point to be decoded belongs may be determined. The target channel group to which the channel corresponding to the feature point to be decoded belongs may include k channels: channel c1, channel c2, ..., and channel ck.

[0142] S4032: Based on at least one feature value of at least one decoded feature point corresponding to a target channel group, and an estimated information matrix corresponding to a target channel group, determine at least one probability distribution parameter corresponding to a feature point to be decoded.

[0143] For example, to determine at least one probability distribution parameter corresponding to a feature point to be decoded, a weighting calculation is performed based on at least one feature value of at least one decoded feature point corresponding to a target channel group, and an estimated information matrix corresponding to the target channel group. For example, at least one decoded feature point corresponding to a target channel group may be at least one decoded feature point corresponding to k channels included within the target channel group.

[0144] S4033: Determines a probability distribution corresponding to a feature point to be decoded based on at least one probability distribution parameter corresponding to a feature point to be decoded.

[0145] For example, after at least one probability distribution parameter corresponding to a feature point to be decoded is determined, the probability distribution corresponding to the feature point to be decoded can be determined based on at least one probability distribution parameter corresponding to the feature point to be decoded.

[0146] For example, at least one probability distribution parameter is a Gaussian distribution parameter, and the determined probability distribution corresponding to the feature point to be decoded is a Gaussian probability distribution.

[0147] S4034: To obtain feature values, decode the feature points to be decoded based on the probability distribution corresponding to the feature points to be decoded.

[0148] For example, in order to obtain the feature value of a feature point to be decoded, the encoded data of the feature value of the feature point to be decoded may be decoded based on a probability distribution corresponding to the feature point to be decoded. In this case, the feature point to be decoded becomes the decoded feature point.

[0149] Therefore, the decoding target feature points of the channels within all target channel groups can be decoded in the manner described above.

[0150] S404: To output the reconstructed image, reconstruction is performed based on C channel feature maps.

[0151] For example, after C channel feature maps are acquired, image reconstruction can be performed based on the C channel feature maps to acquire a reconstructed image.

[0152] Compared to existing techniques in which at least one probability distribution parameter is determined based on the feature value of a decoded feature point of all channels and the estimated information matrix of all channels, in this application, at least one probability distribution parameter corresponding to a feature point to be decoded needs to be determined solely based on at least one feature value of at least one decoded feature point within a feature map of a part of the channel and the estimated information matrix of the part of the channel. This can reduce the computational power required for decoding and improve decoding efficiency.

[0153] Additionally, compared to existing techniques in which context features need to be generated based on at least one feature value of at least one decoded feature point, and then at least one probability distribution parameter is determined based on the context features and the estimated information matrix, in this application, context features do not need to be generated. This further reduces the computational power required for decoding and improves decoding efficiency.

[0154] Figure 5 is a diagram of an example of an end-to-end image compression framework.

[0155] Refer to FIG. 5. For example, an encoder network, a quantization unit (D1), an aggregation unit, a hyper-encoder network, a quantization unit (D2), a hyper-decoder network, a probability estimation unit (V1), and a probability estimation unit (V2) belong to the AI ​​encoding unit in FIG. 1. For example, a decoder network, an aggregation unit, a hyper-decoder network, a probability estimation unit (V1), and a probability estimation unit (V2) belong to the AI ​​decoding unit in FIG. 1.

[0156] For example, the entropy encoding unit (A1) and the entropy encoding unit (B1) belong to the entropy encoding unit in FIG. 1.

[0157] For example, the entropy decoding unit (A2) and the entropy decoding unit (B2) belong to the entropy decoding unit in FIG. 1.

[0158] For example, the AI ​​encoding unit and the AI ​​decoding unit can perform joint training so that the respective networks and units within the AI ​​encoding unit and the AI ​​decoding unit learn about corresponding parameters. For example, the aggregation unit, hyperdecoder network, probability estimation unit (V1), and probability estimation unit (V2) within the AI ​​encoding unit and the aggregation unit, hyperdecoder network, probability estimation unit (V1), and probability estimation unit (V2) within the AI ​​decoding unit can be shared.

[0159] For example, the encoder network may be configured to perform a spatial transformation on the image to be encoded in order to transform the image to be encoded into another space. For example, the encoder network may be a convolutional neural network.

[0160] For example, a hyperencoder network can be configured to extract features. For example, a hyperencoder network can be a convolutional neural network.

[0161] For example, a quantization unit (including a quantization unit (D1) and a quantization unit (D2)) may be configured to perform quantization processing.

[0162] For example, the aggregation unit may be configured to determine at least one probability distribution parameter of the encoding target / decoding target feature point.

[0163] For example, a probability estimation unit (including a probability estimation unit (V1) and a probability estimation unit (V2)) may be configured to estimate a probability and output a probability distribution. Optionally, the probability estimation unit (V1) may be a discrete probability estimation unit such as a multiplication model, and the probability estimation unit (V2) may be a discrete probability estimation unit such as an entropy estimation model.

[0164] For example, the entropy encoding unit (A1) may be configured to perform encoding based on the probability distribution PA1 determined by the probability estimation unit (V1) in order to reduce the statistical redundancy of the output features.

[0165] For example, the entropy encoding unit (B1) may be configured to perform encoding based on a probability distribution PB1 determined by a probability estimation unit (V2) in order to reduce statistical redundancy of output features.

[0166] For example, the entropy decoding unit (A2) may be configured to perform decoding based on the probability distribution PA2 determined by the probability estimation unit (V1).

[0167] For example, the entropy decoding unit (B2) may be configured to perform decoding based on the probability distribution PB2 determined by the probability estimation unit (V2).

[0168] For example, the decoder network can be configured to perform an inverse spatial transformation on the information obtained through entropy decoding and output a reconstructed image. For example, the decoder network can be a convolutional neural network.

[0169] For example, a hyperdecoder network can be configured to process features extracted by a hyperencoder network and output an estimated information matrix. For example, the hyperdecoder network can be a convolutional neural network.

[0170] Refer to Fig. 5. The encoding process may be as follows.

[0171] For example, an image to be encoded is input into an encoder network, and the encoder network transforms the image to be encoded into another space to output a feature map matrix Y1. The feature map matrix Y1 is input into a quantization unit (D1), and the quantization unit (D1) performs quantization processing on the feature map matrix Y1 to output a feature map matrix Y2 (the feature map matrix Y2 includes C channels of feature maps in the above-described embodiment), wherein the feature map matrix Y2 am.

[0172] For example, the quantization unit (D1) can perform quantization processing on the feature values ​​of each feature point in the feature map of each channel in the feature map of the feature map matrix Y1 based on a preset quantization step to obtain the feature map matrix Y2.

[0173] For example, after a feature map matrix Y2 is obtained, in one aspect, the feature map matrix Y2 is input into a hyper-encoder network, and the hyper-encoder network performs feature extraction on the feature map matrix Y2 to obtain a feature map matrix Z1, and then inputs the feature map matrix Z1 to a quantization unit (D2). The quantization unit (D2) performs quantization processing on the feature map matrix Z1 and then outputs the feature map matrix Z2.

[0174] In a possible embodiment, a feature map matrix Z2 is input to a probability estimation unit (V2) and processed by the probability estimation unit (V2) so that the probability distribution PB1 of each feature point within the feature map matrix Z2 is output to an entropy encoding unit (B1). Additionally, the feature map matrix Z2 is input to the entropy encoding unit (B1). The entropy encoding unit (B1) encodes the feature map matrix Z2 based on the probability distribution PB1 and outputs the bitstream SB to the entropy decoding unit (B2). Next, the probability estimation unit (V2) can predict the probability distribution PB2 of the feature points to be decoded within the bitstream SB and input the probability distribution PB2 to the entropy decoding unit (B2). Next, the entropy decoding unit (B2) can decode the feature points to be decoded within the bitstream SB based on the probability distribution PB2 and output the feature map matrix Z2 to a hyperdecoder network. After acquiring the feature map matrix Z2, the hyperdecoder network can process the feature map matrix Z2 to acquire an estimated information matrix of C channels, and input the estimated information matrix of C channels into an aggregation unit.

[0175] In a possible embodiment, the feature map matrix Z2 may be directly input to a hyperdecoder network, the hyperdecoder network processes the feature map matrix Z2 to obtain an estimated information matrix of C channels, and inputs the C estimated information matrix to an aggregation unit. It should be understood that the method of determining the C estimated information matrix in the encoding process is not limited in this application.

[0176] For example, after the feature map matrix Y2 is acquired, in another aspect, the feature map matrix Y2 may be input to an aggregation unit, and the aggregation unit determines at least one probability distribution parameter corresponding to an encoded target feature point corresponding to a target channel group based on at least one feature value of at least one encoded feature point corresponding to a target channel group comprising k channels, and an estimated information matrix corresponding to the target channel group. For this process, refer to the description above. Details are not described again in this specification.

[0177] For example, after obtaining at least one probability distribution parameter corresponding to a feature point to be encoded, the aggregation unit may input at least one probability distribution parameter corresponding to a feature point to be encoded into a probability estimation unit (V1), and the probability estimation unit (V1) determines a probability distribution PA1 corresponding to a feature point to be encoded based on at least one probability distribution parameter corresponding to a feature point to be encoded, and then the probability estimation unit (V1) may input the probability distribution PA1 corresponding to a feature point to be encoded into an entropy encoding unit (A1).

[0178] For example, after the feature map matrix Y2 is obtained, in another aspect, the feature map matrix Y2 can be input to an entropy encoding unit (A1), and the entropy encoding unit (A1) encodes the feature values ​​of the feature points to be encoded within the feature map matrix Y2 based on the probability distribution PA1 corresponding to the feature points to be encoded in order to obtain a bitstream SA. Next, the encoding of the image to be encoded is completed.

[0179] It should be noted that after the encoding of the image to be encoded is completed, both the bitstream SA obtained by encoding the feature map matrix Y2 and the bitstream SB obtained by encoding the feature map matrix Z2 can be transmitted to the decoder side.

[0180] Refer still to Fig. 5. The decoding process may be as follows:

[0181] For example, after receiving bitstream SA and bitstream SB, the decoder side can assign bitstream SA to an entropy decoding unit (A2) for decoding and bitstream SB to an entropy decoding unit (B2) for decoding.

[0182] For example, a probability estimation unit (V2) can predict the probability distribution PB2 of the feature points to be decoded within the bitstream SB and input the probability distribution PB2 to the entropy decoding unit (B2). Next, the entropy decoding unit (B2) can decode the feature points to be decoded within the bitstream SB based on the probability distribution PB2 and output the feature map matrix Z2 to the hyperdecoder network. After obtaining the feature map matrix Z2, the hyperdecoder network can process the feature map matrix Z2 to obtain an estimated information matrix of C channels and input the estimated information matrix of C channels to the aggregation unit.

[0183] For example, the bitstream SA contains encoded data of feature values ​​of each feature point in the feature map matrix Y2, and the entropy decoding unit (A2) decodes the encoded data of feature values ​​of each feature point in the bitstream SA to obtain feature values ​​corresponding to each feature point, thereby obtaining the feature map matrix Y2.

[0184] For example, for one feature point to be decoded, the entropy decoding unit (A2) may input at least one feature value corresponding to at least one decoded feature point to the aggregation unit, and the aggregation unit determines a target channel group to which the channel corresponding to the feature point to be decoded belongs, and determines at least one probability distribution parameter corresponding to the feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to the target channel group and an estimated information matrix corresponding to the target channel group. For details, refer to the description above. Details are not described again in this specification. Next, at least one probability distribution parameter corresponding to the feature point to be decoded is output to the probability estimation unit (V1). Next, the probability estimation unit (V1) performs probability estimation based on at least one probability distribution parameter corresponding to the feature point to be decoded, predicts the probability distribution PA2 corresponding to the feature point to be decoded, and inputs the probability distribution PA2 corresponding to the feature point to be decoded to the entropy decoding unit (A2). Next, the entropy encoding unit (A2) can decode the encoded data of the feature value of the feature point to be decoded based on the probability distribution PA2 corresponding to the feature point to be decoded in order to obtain the feature value. In this case, the above steps are repeated. The entropy decoding unit (A2) can decode the bitstream SA and output the feature map matrix Y2 to the decoder network, and the decoder network performs an inverse spatial transformation on the feature map matrix Y2 to obtain the reconstructed image.

[0185] For example, the entropy encoding unit (A1) may perform parallel encoding or serial encoding on feature points to be encoded in different target channel groups. This is not limited in this application. Correspondingly, the entropy decoding unit (A2) may perform parallel decoding or serial decoding on feature points to be decoded in different target channel groups. This is not limited in this application.

[0186] It should be noted that in the encoding process, the feature map matrix Y1 can alternatively be input to a hyperencoder network, and the feature map matrix Z2 is obtained through the hyperencoder network and the quantization unit (D2). This is not limited to this application.

[0187] It should be noted that the network or unit within the right dashed-line box in FIG. 5 may alternatively be a different network or a different unit and may be specifically configured based on requirements. This is not limited to this application.

[0188] It should be noted that networks and units configured to generate different estimated information matrices of at least one probability distribution parameter may be further included within the AI ​​encoding unit, AI decoding unit, entropy encoding unit, and entropy decoding unit in this application. Next, the other estimated information matrix is ​​input to an aggregation unit. The aggregation unit determines at least one probability distribution parameter of the encoding target / decoding target feature point based on the feature value of the encoded / decoded feature point corresponding to a target channel group comprising k channels, the estimated information matrix corresponding to the target channel group, and the other estimated information matrix. This is not limited in this application.

[0189] For example, the AI ​​encoding unit and the AI ​​decoding unit (excluding the aggregation unit and the probability estimation unit (V1)) may be placed within an NPU (Neural Network Processing Unit) or a GPU (Graphics Processing Unit). For example, the entropy encoding unit, the entropy decoding unit, the aggregation unit, and the probability estimation unit (V1) may be placed within a CPU (Central Processing Unit). Therefore, compared to existing technology in which the aggregation unit and probability estimation unit (V1) are deployed within the GPU, in this application, during the decoding process, whenever the CPU obtains a feature value of a decoded feature point through decoding, the feature value is directly stored in the CPU's memory, and the aggregation unit and probability estimation unit (V1) within the CPU determine at least one probability distribution parameter of the feature point to be decoded, thereby avoiding frequent communication between the CPU and the GPU during the decoding process and improving decoding efficiency.

[0190] In a possible embodiment, among the estimated information matrices of C channels, the estimated information matrix output by the hyperdecoder network of FIG. 5 includes a first feature matrix of C channels and a second feature matrix of C channels. The first feature matrix of C channels may be used to determine a first probability distribution parameter, and the second feature matrix of C channels may be used to determine a second probability distribution parameter. The first feature matrix may include first features of H*W feature points, and the second feature matrix may include second features of H*W feature points. In this case, the process of determining at least one probability distribution parameter of the feature point to be encoded may be as follows.

[0191] Figure 6 is a diagram of an example of an encoding process.

[0192] S601: Acquired the image to be encoded.

[0193] S602: Generate a feature map of C channels based on an image to be encoded, wherein the feature map contains feature values ​​of multiple feature points, and C is a positive integer.

[0194] S603: Generate an estimated information matrix of C channels based on a feature map of C channels, wherein the estimated information matrix contains estimated information of multiple feature points.

[0195] S604: Groups C channels into N channel groups.

[0196] For example, for S601 to S604, refer to the descriptions of S201 to S204.

[0197] The estimated information matrix of C channels includes a first feature matrix of C channels and a second feature matrix of C channels. The first feature matrix of C channels can be used to determine a first probability distribution parameter, and the second feature matrix of C channels can be used to determine a second probability distribution parameter. The first feature matrix may include first features of H*W feature points, and the second feature matrix may include second features of H*W feature points.

[0198] S605: Based on at least one feature value of at least one encoded feature point corresponding to a target channel group, and a first feature matrix corresponding to a target channel group, determine at least one first probability distribution parameter corresponding to an encoded target feature point corresponding to a target channel group.

[0199] For example, regarding S605, in order to determine at least one first probability distribution parameter of an encoding target feature point corresponding to a target channel group, refer to the description of S205 above. Details are not described again in this specification.

[0200] S606: Based on at least one feature value of at least one encoded feature point corresponding to a target channel group, and a second feature matrix corresponding to a target channel group, determine at least one second probability distribution parameter corresponding to an encoded target feature point corresponding to a target channel group.

[0201] For example, regarding S606, to determine at least one second probability distribution parameter of an encoding target feature point corresponding to a target channel group, refer to the description of S205 above. Details are not described again in this specification.

[0202] For example, at least one probability distribution parameter of an encoding target feature point corresponding to a target channel group may include a first probability distribution parameter of an encoding target feature point corresponding to a target channel group, and at least one second probability distribution parameter of an encoding target feature point corresponding to a target channel group.

[0203] In a possible implementation, the first probability distribution parameter is the mean, and the second probability distribution parameter is the variance.

[0204] In a possible implementation, the first probability distribution parameter is the variance, and the second probability distribution parameter is the mean.

[0205] S607: Determine a probability distribution corresponding to a feature point to be encoded based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to a feature point to be encoded.

[0206] S608: To obtain a bitstream, encoding target feature points based on a probability distribution corresponding to the encoding target feature points.

[0207] For example, regarding S607 and S608, refer to the descriptions of S206 and S207 above. Details are not described again in this specification.

[0208] FIG. 7 is a diagram of an example of a decoding process. The decoding process in the embodiment of FIG. 7 corresponds to the encoding process in the embodiment of FIG. 6.

[0209] S701: Received bitstream.

[0210] For example, the bitstream received by the decoder side may include the image bitstream and the bitstream corresponding to the image among the bitstreams of previous information of C channels.

[0211] For example, the image bitstream may contain encoded data of feature values ​​of feature points of C channels, and the bitstream of previous information of C channels may contain encoded data of previous information of C channels.

[0212] S702: Decode the bitstream to obtain the estimated information matrix of C channels.

[0213] For example, after a bitstream is received, the bitstream may be parsed to obtain the encoded data of the previous information of C channels. Next, to obtain the first feature matrix of C channels and the second feature matrix of C channels, entropy decoding and hyper-decoding are performed on the encoded data of the previous information of C channels.

[0214] S703: Decodes the bitstream to obtain feature values ​​of feature points of C channels based on the estimated information matrix of C channels, in order to obtain feature maps of C channels.

[0215] For example, regarding S703, refer to the description of S403 above. Details are not described further in this specification.

[0216] For example, the feature point to be decoded can be determined from the feature point that has not been decoded in the decoding order corresponding to the encoding order in FIG. 3c. Next, the feature value of the feature point to be decoded can be obtained through decoding with reference to the following steps (7031 to 7035).

[0217] S7031: From N channel groups obtained by grouping C channels, determine the target channel group to which the channel corresponding to the feature point to be decoded belongs.

[0218] S7032: Based on at least one feature value of at least one decoded feature point corresponding to a target channel group, and a first feature matrix corresponding to a target channel group, determine at least one first probability distribution parameter corresponding to a feature point to be decoded.

[0219] S7033: Based on at least one feature value of at least one decoded feature point corresponding to a target channel group, and a second feature matrix corresponding to a target channel group, determine at least one second probability distribution parameter corresponding to a feature point to be decoded.

[0220] S7034: Determines a probability distribution corresponding to a feature point to be decoded based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to a feature point to be decoded.

[0221] S7035: To obtain feature values, decode the feature points to be decoded based on the probability distribution corresponding to the feature points to be decoded.

[0222] For example, regarding S7031 to S7035, refer to the descriptions of S4031 to S4034 above. Details are not described again in this specification.

[0223] S704: To output the reconstructed image, reconstruction is performed based on C channel feature maps.

[0224] For example, regarding S704, refer to the description of S404 above. Details are not described further in this specification.

[0225] In a possible embodiment, among the estimated information matrices of C channels, the estimated information matrix output by the hyperdecoder network of FIG. 5 includes a first feature matrix of C channels and a second probability distribution parameter matrix of C channels. The first feature matrix of C channels may be used to determine the first probability distribution parameter, and the first feature matrix may include first features of H*W feature points. The second probability distribution parameter matrix may include second probability distribution parameters of H*W feature points. In this case, the process of determining at least one probability distribution parameter of the feature point to be encoded may be as follows.

[0226] Figure 8 is a diagram of an example of an encoding process.

[0227] S801: Acquired the image to be encoded.

[0228] S802: Generate a feature map of C channels based on an image to be encoded, wherein the feature map contains feature values ​​of multiple feature points, and C is a positive integer.

[0229] S803: Generate an estimated information matrix of C channels based on a feature map of C channels, wherein the estimated information matrix contains estimated information of multiple feature points.

[0230] S804: Groups C channels into N channel groups.

[0231] For example, for S801 to S804, refer to the descriptions of S201 to S204.

[0232] The estimated information matrix of C channels includes a first feature matrix of C channels and a second probability distribution parameter matrix of C channels. The first feature matrix of C channels can be used to determine the first probability distribution parameter, and the first feature matrix may include first features of H*W feature points. The second probability distribution parameter matrix may include second probability distribution parameters of H*W feature points.

[0233] S805: Based on at least one feature value of at least one encoded feature point corresponding to a target channel group, and a first feature matrix corresponding to a target channel group, determine at least one first probability distribution parameter corresponding to an encoded target feature point corresponding to a target channel group.

[0234] For example, regarding S805, to determine at least one first probability distribution parameter of an encoding target feature point corresponding to a target channel group, refer to the description of S205 above. Details are not described again in this specification.

[0235] In a possible implementation, the first probability distribution parameter is the mean, and the second probability distribution parameter is the variance.

[0236] In a possible implementation, the first probability distribution parameter is the variance, and the second probability distribution parameter is the mean.

[0237] S806: Based on a second probability distribution parameter matrix corresponding to a target channel group, determine at least one second probability distribution parameter of an encoding target feature point corresponding to a target channel group.

[0238] For example, one encoding target feature point of channel c (where c is a positive integer less than or equal to C) within a target channel group is used as an example for explanation. For example, the second probability distribution parameter matrix of channel c can be determined from the second probability distribution parameter matrices of C channels. Next, at least one second probability distribution parameter corresponding to the encoding target feature point is determined from the second probability distribution parameter matrix of channel c based on the location of the encoding target feature point within the feature map of channel c.

[0239] S807: Determine a probability distribution corresponding to a feature point to be encoded based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to a feature point to be encoded.

[0240] S808: To obtain a bitstream, encoding target feature points based on a probability distribution corresponding to the encoding target feature points.

[0241] For example, regarding S807 and S808, refer to the descriptions of S206 and S207 above. Details are not described again in this specification.

[0242] FIG. 9 is a diagram of an example of a decoding process. The decoding process in the embodiment of FIG. 9 corresponds to the encoding process in the embodiment of FIG. 8.

[0243] S901: Received bitstream.

[0244] For example, regarding S901, refer to the description of S701 above. Details are not described further in this specification.

[0245] S902: Decodes the bitstream to obtain the estimated information matrix of C channels.

[0246] For example, after a bitstream is received, the bitstream may be parsed to obtain encoded data of previous information of C channels. Next, entropy decoding and hyper-decoding are performed on the encoded data of previous information of C channels to obtain a first feature matrix of C channels and a second probability distribution parameter matrix of C channels.

[0247] S903: Decodes the bitstream to obtain feature values ​​of feature points of C channels based on the estimated information matrix of C channels, in order to obtain feature maps of C channels.

[0248] For example, regarding S903, refer to the description of S403 above. Details are not described further in this specification.

[0249] For example, the feature point to be decoded can be determined from the feature point that has not been decoded in the decoding order corresponding to the encoding order in FIG. 3c. Next, the feature value of the feature point to be decoded can be obtained through decoding by referring to the following steps (S9031 to S9034).

[0250] S9031: From N channel groups obtained by grouping C channels, determine the target channel group to which the channel corresponding to the feature point to be decoded belongs.

[0251] S9032: Based on at least one feature value of at least one decoded feature point corresponding to a target channel group, and a first feature matrix corresponding to a target channel group, determine at least one first probability distribution parameter corresponding to a feature point to be decoded.

[0252] S9033: Determine at least one second probability distribution parameter of a feature point to be decoded based on a second probability distribution parameter matrix corresponding to a target channel group.

[0253] For example, it is assumed that the channel corresponding to the feature point to be decoded is channel c. In this case, the second probability distribution parameter matrix of channel c can be determined from the second probability distribution parameter matrices of C channels. Next, at least one second probability distribution parameter corresponding to the feature point to be decoded is determined from the second probability distribution parameter matrix of channel c based on the location of the feature point to be decoded.

[0254] S9034: Determines a probability distribution corresponding to a feature point to be decoded based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to a feature point to be decoded.

[0255] S9035: To obtain feature values, decode the feature points to be decoded based on the probability distribution corresponding to the feature points to be decoded.

[0256] For example, regarding S9031 to S9035, refer to the descriptions of S4031 to S4034 above. Details are not described again in this specification.

[0257] S904: To output the reconstructed image, image reconstruction is performed based on C channel feature maps.

[0258] For example, regarding S904, refer to the description of S404 above. Details are not described further in this specification.

[0259] In a possible embodiment, among the estimated information matrices of C channels, the estimated information matrix output by the hyperdecoder network of FIG. 5 includes a first probability distribution parameter matrix of C channels and a second feature matrix of C channels. The second feature matrix of C channels may be used to determine the second probability distribution parameter, and the second feature matrix may include second features of H*W feature points. The first probability distribution parameter matrix may include first probability distribution parameters of H*W feature points. In this case, the process of determining at least one probability distribution parameter of the feature point to be encoded may be as follows.

[0260] Figure 10 is a diagram of an example of an encoding process.

[0261] S1001: Acquire the image to be encoded.

[0262] S1002: Generate a feature map of C channels based on an image to be encoded, wherein the feature map contains feature values ​​of multiple feature points, and C is a positive integer.

[0263] S1003: Generate an estimated information matrix of C channels based on a feature map of C channels, wherein the estimated information matrix contains estimated information of multiple feature points.

[0264] S1004: Group C channels into N channel groups.

[0265] For example, for S1001 to S1004, refer to the descriptions of S201 to S204.

[0266] The estimated information matrix of C channels includes a first probability distribution parameter matrix of C channels and a second feature matrix of C channels. The second feature matrix of C channels can be used to determine the second probability distribution parameter, and the second feature matrix may include second features of H*W feature points. The first probability distribution parameter matrix may include the first probability distribution parameter of H*W feature points.

[0267] S1005: Based on at least one feature value of at least one encoded feature point corresponding to a target channel group, and a second feature matrix corresponding to a target channel group, determine at least one second probability distribution parameter corresponding to an encoded target feature point corresponding to a target channel group.

[0268] For example, regarding S1005, to determine at least one second probability distribution parameter of an encoding target feature point corresponding to a target channel group, refer to the description of S205 above. Details are not described again in this specification.

[0269] In a possible implementation, the first probability distribution parameter is the mean, and the second probability distribution parameter is the variance.

[0270] In a possible implementation, the first probability distribution parameter is the variance, and the second probability distribution parameter is the mean.

[0271] S1006: Based on a first probability distribution parameter matrix corresponding to a target channel group, determine at least one first probability distribution parameter of an encoding target feature point corresponding to a target channel group.

[0272] For example, one encoding target feature point of channel c within the target channel group is used as an example for explanation. For example, the first probability distribution parameter matrix of channel c can be determined from the first probability distribution parameter matrices of C channels. Next, at least one first probability distribution parameter corresponding to the encoding target feature point is determined from the first probability distribution parameter matrix of channel c based on the location of the encoding target feature point within the feature map of channel c.

[0273] S1007: Determine a probability distribution corresponding to an encoding target feature point based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to an encoding target feature point.

[0274] S1008: To obtain a bitstream, encoding target feature points based on a probability distribution corresponding to the encoding target feature points.

[0275] For example, regarding S1007 and S1008, refer to the descriptions of S206 and S207 above. Details are not described again in this specification.

[0276] FIG. 11 is a diagram of an example of a decoding process. The decoding process in the embodiment of FIG. 11 corresponds to the encoding process in the embodiment of FIG. 10.

[0277] S1101: Received bitstream.

[0278] For example, regarding S1101, refer to the description of S701 above. Details are not described further in this specification.

[0279] S1102: Decode the bitstream to obtain the estimated information matrix of C channels.

[0280] For example, after a bitstream is received, the bitstream may be parsed to obtain the encoded data of the previous information of C channels. Next, entropy decoding and hyper-decoding are performed on the encoded data of the previous information of C channels to obtain the first probability distribution parameter matrix of C channels and the second feature matrix of C channels.

[0281] S1103: Decodes the bitstream to obtain feature values ​​of feature points of C channels based on the estimated information matrix of C channels, in order to obtain feature maps of C channels.

[0282] For example, regarding S1103, refer to the description of S403 above. Details are not described further in this specification.

[0283] For example, the feature point to be decoded can be determined from the feature point that has not been decoded in the decoding order corresponding to the encoding order in FIG. 3c. Next, the feature value of the feature point to be decoded can be obtained through decoding by referring to the following steps (S11031 to S11034).

[0284] S11031: From N channel groups obtained by grouping C channels, determine the target channel group to which the channel corresponding to the feature point to be decoded belongs, wherein the channel group includes k channels.

[0285] S11032: Based on at least one feature value of at least one decoded feature point corresponding to a target channel group, and a second feature matrix corresponding to a target channel group, determine at least one second probability distribution parameter corresponding to a feature point to be decoded.

[0286] S11033: Based on a first probability distribution parameter matrix corresponding to a target channel group, determine at least one first probability distribution parameter of a feature point to be decoded.

[0287] For example, it is assumed that the channel corresponding to the feature point to be decoded is channel c (where c is a positive integer less than or equal to C). In this case, the first probability distribution parameter matrix of channel c can be determined from the first probability distribution parameter matrices of C channels. Next, at least one first probability distribution parameter corresponding to the feature point to be decoded is determined from the first probability distribution parameter matrix of channel c based on the spatial location of the feature point to be decoded.

[0288] S11034: Determines a probability distribution corresponding to a feature point to be decoded based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to a feature point to be decoded.

[0289] S11035: To obtain feature values, decode the feature points to be decoded based on the probability distribution corresponding to the feature points to be decoded.

[0290] For example, regarding S11031 to S11035, refer to the descriptions of S4031 to S4034 above. Details are not described again in this specification.

[0291] S1104: To output the reconstructed image, image reconstruction is performed based on C channel feature maps.

[0292] For example, regarding S1104, refer to the description of S404 above. Details are not described further in this specification.

[0293] The following explains, by using an example, how an aggregation unit determines at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group. The explanation is provided by using an example where the first probability distribution parameter is the mean and the second probability distribution parameter is the variance.

[0294] For example, to determine at least one probability distribution parameter corresponding to an encoded target feature point corresponding to a target channel group, linear weighting may be performed on at least one feature value of at least one encoded feature point corresponding to a target channel group including k channels, and on an estimated information matrix corresponding to the target channel group. Therefore, the amount of computation required to determine the probability distribution parameter can be reduced from the original tens of thousands of multiplication-accumulation calculations to at least a few multiplication-accumulation calculations and up to a few hundred multiplication-accumulation calculations. This significantly reduces the amount of computation.

[0295] For example, the size of the linear weighting window may be preset, for example, ks1*ks2. Next, to determine at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group, linear weighting is performed on at least one feature value of at least one encoded feature point within the linear weighting window and within a feature map corresponding to the target channel group, and on the estimated information of a feature point corresponding to a part of a location within the linear weighting window and within an estimated information matrix corresponding to the target channel group, wherein ks1 and ks2 are integers greater than 1, and ks1 and ks2 may be equal or not equal. This is not limited in this application. Therefore, at least one probability distribution parameter corresponding to an encoding target feature point is determined based on the feature value of a feature point around the encoding target feature point and the estimated information, so that the amount of computation for the probability distribution parameter can be reduced while ensuring the accuracy of the determined probability distribution parameter.

[0296] For example, the following embodiments in FIGS. 12a to 12d and the following embodiments in FIGS. 13a and 13b describe a process in which linear weighting is performed on at least one feature value of at least one encoded feature point corresponding to a target channel group including k channels, and on a first feature matrix corresponding to a target channel group, in order to determine a first probability distribution parameter corresponding to an encoded target feature point corresponding to a target channel group.

[0297] FIG. 12a is a diagram of an example of a process for determining probability distribution parameters. In the embodiment of FIG. 12a, the number of channels k included in the target channel group is 1, ks1 = ks3 = 3, and the encoding order on the encoder side is illustrated in (1) of FIG. 3c. The target channel group is assumed to include a first channel (hereinafter referred to as channel c1). In the feature map of channel c1, gray blocks represent encoded feature points, white blocks represent unencoded feature points, and blocks filled with diagonal lines represent feature points D to be encoded. In the first feature matrix of channel c1, gray blocks represent encoded locations, white blocks represent unencoded locations, and blocks filled with diagonal lines represent locations L to be encoded. The encoded location is the location of the encoded feature point, the unencoded location is the location of the unencoded feature point, and the encoding target location L is the location of the encoding target feature point D.

[0298] Refer to FIG. 12a. For example, a first target region Q1, located within the feature map of channel c1 and centered on the encoding target feature point D, can be determined based on the size of a linear weighting window. A second target region Q2, located within the first feature matrix of channel c1 and centered on the encoding target location L, is determined based on the size of a linear weighting window. The sizes of the first target region Q1 and the second target region Q2 are the same, and both are ks1*ks2.

[0299] Refer to FIG. 12a. For example, an encoded feature point within a first target region Q1 can be determined as a feature point for linear weighting calculation within a feature map, that is, a feature point corresponding to the gray block within the first target region Q1 in FIG. 12a. The location of the encoding target feature point D is assumed to be (c1, w, h) (where c1 is an integer from 1 to C, w is an integer from 1 to W, and h is an integer from 1 to H). In this case, the locations of the feature points for linear weighting calculation within the feature map of channel c1 include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1).

[0300] For example, a first target location for linear weighting calculation within a first feature matrix of channel c1 may be determined based on a location other than an encoded location within a second target region Q2. In a possible embodiment, an encoded target location L within the second target region Q2 and at least one other unencoded location (hereinafter referred to as the first unencoded location) may be determined as the first target location. The first unencoded location is a location other than the encoded target location L among all unencoded locations within the second target region Q2.

[0301] Refer to FIG. 12a. In a possible embodiment, locations other than the encoded locations within the second target area Q2 (i.e., the encoded target location L and all first unencoded locations within the second target area Q2) may be determined as locations within the second target area Q2 in FIG. 12a that correspond to the first target locations, i.e., blocks filled with diagonal lines and blocks filled with dots.

[0302] It is assumed that the location of the encoding target feature point D is (c1, w, h), and the locations of the feature points for linear weighting calculation within the feature map of channel c1 include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1). In this case, the first target locations include (c1, w, h), (c1, w+1, h), (c1, w-1, h+1), (c1, w, h+1), and (c1, w+1, h+1).

[0303] For example, in the training process, the aggregation unit learns a weight matrix (i.e., a preset weight matrix) that corresponds to each channel and is used for linear weighting. The preset weight matrix may contain weight maps for k channels, and the size of the weight map is equal to the size of the linear weighting window, e.g., ks1*ks2. In the embodiment of FIG. 12a, the preset weight matrix corresponding to channel c1 contains a weight map for one channel, and the weight map corresponding to channel c1 contains 9 weights: ω1 , ω2 , ω3 , ω4 , ω5 , ω6 , ω7 , 8 hours , and oh9 Includes

[0304] For example, to obtain at least one first probability distribution parameter corresponding to an encoding target feature point D, linear weighting may be performed on the feature value of an encoded feature point within a first target region Q1 and the first feature of a feature point corresponding to a first target location within a second target region Q2, based on a weight map corresponding to channel c1.

[0305] The location of the encoding target feature point D is (c1, w, h), and the feature values ​​of the encoded feature points within the first target region Q1 are represented as y[c1][w-1][h], y[c1][w-1][h-1], y[c1][w][h-1], and y[c1][w+1][h-1], and the first feature of the feature point corresponding to the first target location within the second target region Q2 is and It is assumed to be represented as. In this case, at least one first probability distribution parameter = corresponding to the encoding target feature point D in the feature map of channel c1.

[0306]

[0307] FIG. 12b is a diagram of an example of a process for determining probability distribution parameters. In the embodiment of FIG. 12b, the number of channels k included in the target channel group is 1, ks1 = ks3 = 3, and the encoding order on the encoder side is illustrated in (1) of FIG. 3c. The target channel group is assumed to include a first channel (hereinafter referred to as channel c1). In the feature map of channel c1, gray blocks represent encoded feature points, white blocks represent unencoded feature points, and blocks filled with diagonal lines represent feature points D to be encoded. In the first feature matrix of channel c1, gray blocks represent encoded locations, white blocks represent unencoded locations, and blocks filled with diagonal lines represent locations L to be encoded. The encoded location is the location of the encoded feature point, the unencoded location is the location of the unencoded feature point, and the encoding target location L is the location of the encoding target feature point D.

[0308] For example, the difference between FIG. 12b and FIG. 12a is that the method of determining the first target location for linear weighting calculation in the first feature matrix in FIG. 12b, based on a location other than the encoded location within the second target region Q2, is different from the method in FIG. 12a.

[0309] Refer to FIG. 12b. In a possible embodiment, an encoding target location L within a second target region Q2 can be determined as a first target location. The location of the encoding target feature point D is (c1, w, h), and the locations of the feature points for linear weighting calculation within the feature map of channel c1 are assumed to include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1). In this case, the first target location is (c1, w, h).

[0310] In this case, in the weight map corresponding to channel c1, ω6 , ω7 , 8 hours , and oh9 All are 0. In this case, at least one first probability distribution parameter = corresponding to the encoding target feature point D in the feature map of channel c1

[0311]

[0312] A test set including multiple images having different resolutions is used below for testing to compare the encoding performance of the encoding based on the embodiment in FIG. 12a with the encoding performance of the encoding based on the embodiment in FIG. 12b.

[0313] For example, to obtain the rate gain (BD-rate, where a larger negative value of the BD-rate indicates higher encoding performance corresponding to the encoding based on the embodiment in FIG. 12a) obtained by comparing the encoding based on the embodiment in FIG. 12a with the encoding in the prior art, and the BD-rate obtained by comparing the encoding based on the embodiment in FIG. 12b with the encoding in the prior art, 16 encoding target images having different resolutions are used for testing. Details may be illustrated in Table 1:

[0314]

[0315] Refer to Table 1. The encoding performance of the encoding based on the embodiment in FIG. 12a is better than the encoding performance of the encoding based on the embodiment in FIG. 12b on the three channels of Y, U, and V. In other words, for the same encoding quality, the bit rate corresponding to the encoding based on the embodiment in FIG. 12a is smaller than the bit rate corresponding to the encoding based on the embodiment in FIG. 12b.

[0316] FIG. 12c is a diagram of an example of a process for determining probability distribution parameters. In the embodiment of FIG. 12c, the number of channels k included in the target channel group is 1, ks1 = ks3 = 3, and the encoding order on the encoder side is shown in (1) of FIG. 3c. The target channel group is assumed to include a first channel (hereinafter referred to as channel c1). In the feature map of channel c1, gray blocks represent encoded feature points, white blocks represent unencoded feature points, and blocks filled with diagonal lines represent feature points D to be encoded. In the first feature matrix of channel c1, gray blocks represent encoded locations, white blocks represent unencoded locations, and blocks filled with diagonal lines represent locations L to be encoded. The encoded location is the location of the encoded feature point, the unencoded location is the location of the unencoded feature point, and the encoding target location L is the location of the encoding target feature point D.

[0317] For example, the difference between FIG. 12c and FIG. 12a is that the method of determining the first target location for linear weighting calculation in the first feature matrix in FIG. 12c, based on a location other than the encoded location within the second target region Q2, is different from the method in FIG. 12a.

[0318] Refer to FIG. 12c. For example, an encoding target location L and a first unencoded location within a second target region Q2 are determined as the first target location. The location of the encoding target feature point D is (c1, w, h), and it is assumed that the locations of the feature points for linear weighting calculation within the feature map of channel c1 include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1). In this case, the first target locations are (c1, w, h) and (c1, w+1, h).

[0319] In this case, in the weight map corresponding to channel c1, ω7 , 8 hours , and oh9 All are 0. In this case, at least one first probability distribution parameter = corresponding to the encoding target feature point D in the feature map of channel c1

[0320]

[0321] It should be understood that the location of the feature point D to be encoded is (c1, w, h), and the locations of the feature points for linear weighting calculation within the feature map are assumed to include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1). In this case, in a possible embodiment, the first target position may be (c1, w, h), (c1, w+1, h), and (c1, w-1, h+1); in a possible embodiment, the first target position may be (c1, w, h), (c1, w+1, h), (c1, w-1, h+1), and (c1, w, h+1); in a possible embodiment, the first target position may be (c1, w, h) and (c1, w-1, h+1); and in a possible embodiment, the first target position may be (c1, w, h) and (c1, w, h+1), etc. This is not limited to the present application.

[0322] FIG. 12d is a diagram of an example of a process for determining probability distribution parameters. In the embodiment of FIG. 12d, the number of channels k included in the target channel group is 2, the target channel group includes 2 channels: a first channel (hereinafter referred to as channel c1) and a second channel (hereinafter referred to as channel c2), ks1 = ks3 = 3, and the encoding order on the encoder side is shown in (1) in FIG. 3c.

[0323] In the feature map of channel 1 and the feature map of channel c2, gray blocks represent encoded feature points, and white blocks represent unencoded feature points. In the feature map of channel c1, blocks filled with diagonal lines represent feature points D1 to be encoded. In the feature map of channel c2, blocks filled with diagonal lines indicate feature points D2 corresponding to the location of feature points D1 to be encoded.

[0324] In the first feature matrix of channel 1 and the first feature matrix of channel c2, gray blocks represent encoded locations, and white blocks represent unencoded locations. Blocks filled with diagonal lines within the first feature matrix of channel c1 represent the encoding target location L1, and blocks filled with diagonal lines within the first feature matrix of channel c2 represent the location L2. The encoded location is the location of the encoded feature point, and the unencoded location is the location of the unencoded feature point; the encoding target location L1 is the location of the encoding target feature point D1, and the location L2 is the location corresponding to the encoding target location L1 (i.e., the location of the feature point D2).

[0325] Refer to FIG. 12d. For example, a first target region Q11, located within the feature map of channel c1 and centered on the feature point D1 to be encoded, can be determined based on the size of the linear weighting window, and a first target region Q12, located within the feature map of channel c2 and centered on the feature point corresponding to the location of the feature point D1 to be encoded, is determined based on the size of the linear weighting window. It is assumed that the location of the feature point D1 to be encoded within the feature map of channel c1 is (c1, w, h). In this case, the feature point D2, whose location is (c2, w, h) within the feature map of channel c2, is the feature point corresponding to the location of the feature point D1 to be encoded.

[0326] Refer to FIG. 12d. For example, a second target region Q21, which is located within the first feature matrix of channel c1 and is used as the center of the encoding target location L1, can be determined based on the size of the linear weighting window, and a second target region Q22, which is located within the first feature matrix of channel c2 and is used as the center of the location corresponding to the encoding target location L1 (i.e., location L2), is determined based on the size of the linear weighting window.

[0327] The sizes of the first target region Q11, the first target region Q12, the second target region Q21, and the second target region Q22 are the same, and they are all ks1*ks2.

[0328] Refer to FIG. 12d. For example, an encoded feature point within a first target region Q11 within a feature map of channel c1 can be determined as a feature point for linear weighting calculation within the feature map of channel c1, that is, a feature point corresponding to the gray block within the first target region Q11 in FIG. 12d. The location of the encoding target feature point D1 is assumed to be (c1, w, h). In this case, the locations of the feature points for linear weighting calculation within the feature map of channel c1 include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1).

[0329] Refer to FIG. 12d. For example, an encoded feature point within a first target region Q12 within a feature map of channel c2 can be determined as a feature point for linear weighting calculation within the feature map of channel c2, that is, a feature point corresponding to the gray block within the first target region Q12 in FIG. 12d. It is assumed that the location of the encoding target feature point D1 is (c1, w, h). In this case, the location of feature point D2 is (c1, w, h), and the locations of the feature points for linear weighting calculation within the feature map of channel c2 include (c2, w-1, h), (c2, w-1, h-1), (c2, w, h-1), and (c2, w+1, h-1).

[0330] For example, a first target location for linear weighting calculation within a first feature matrix of channel c1 may be determined based on a location other than an encoded location within a second target region Q21 within the first feature matrix of channel c1. In a possible embodiment, an encoded target location L1 within the second target region Q21 and at least one first unencoded location may be determined as the first target location. The first unencoded location is a location other than an encoded target location among all unencoded locations within the second target region Q21.

[0331] Refer to FIG. 12d. Locations other than the encoded locations within the second target region Q21 in the first feature matrix of channel c1 (i.e., the encoding target location L1 and all first unencoded locations within the second target region Q21) can be determined as first target locations. The location of the encoding target feature point D1 is (c1, w, h), and it is assumed that the locations of the feature points for linear weighting calculation within the first feature matrix of channel c1 include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1). In this case, the first target locations include (c1, w, h), (c1, w+1, h), (c1, w-1, h+1), (c1, w, h+1), and (c1, w+1, h+1).

[0332] For example, a first target location for linear weighting calculation within a first feature matrix of channel c2 may be determined based on a location other than an encoded location within a second target region Q22 within the first feature matrix of channel c2. In a possible embodiment, a location L2 corresponding to an encoded target location within the second target region Q22 and at least one other unencoded location (hereinafter referred to as the second unencoded location) may be determined as the first target location. The second unencoded location is a location other than location L2 among all unencoded locations within the second target region Q22.

[0333] Refer to FIG. 12d. Locations other than the encoded locations within the second target region Q22 in the first feature matrix of channel 2 (i.e., location L2 corresponding to the encoding target location and located within the second target region Q22, and all second unencoded locations within the second target region Q22) can be determined as the first target locations. The location of the encoding target feature point D1 is (c1, w, h), and it is assumed that the locations of the feature points for linear weighting calculation within the first feature matrix of channel c2 include (c2, w-1, h), (c2, w-1, h-1), (c2, w, h-1), and (c2, w+1, h-1). In this case, the first target locations include (c2, w, h), (c2, w+1, h), (c2, w-1, h+1), (c2, w, h+1), and (c2, w+1, h+1).

[0334] For example, a preset weight matrix corresponding to channel c1 may be determined, and the preset weight matrix may include a weight map (11) and a weight map (12), as illustrated in FIG. 12d. The size of the weight map (11) is 3*3, and the weight map (11) includes nine weights: ω11, ω12, ω13, ω14, ω15, ω16, ω17, ω18, and ω19. The size of the weight map (12) is 3*3, and the weight map (12) includes nine weights: ω21, ω22, ω23, ω24, ω25, ω26, ω27, ω28, and ω29.

[0335] For example, to obtain at least one first probability distribution parameter corresponding to an encoded target feature point D1 within a feature map of channel c1, linear weighting may be performed on a feature value of an encoded feature point within a first target region Q11 within a feature map of channel c1 and a first feature of a feature point corresponding to a first target location within a second target region Q21 within a first feature matrix of channel c1, based on a weight map (11) corresponding to channel c1; and linear weighting may be performed on a feature value of an encoded feature point within a first target region Q12 within a feature map of channel c2 and a first feature of a feature point corresponding to a first target location within a second target region Q22 within a first feature matrix of channel c2, based on a weight map (12) corresponding to channel c1.

[0336] The feature values ​​of the encoded feature points within the first target region Q11 in the feature map of channel c1 are represented as y[c1][w-1][h], y[c1][w-1][h-1], y[c1][w][h-1], and y[c1][w+1][h-1], and the first features of the feature points corresponding to the first target positions within the second target region Q21 in the first feature matrix of channel c1 are and Represented as such, the feature values ​​of the encoded feature points within the first target region Q12 in the feature map of channel c2 are represented as y[c2][w-1][h], y[c2][w-1][h-1], y[c2][w][h-1], and y[c2][w+1][h-1], and the first features of the feature points corresponding to the first target positions within the second target region Q22 in the first feature matrix of channel c2 are and It is assumed to be represented as. In this case, at least one first probability distribution parameter = corresponding to the encoding target feature point D1 in the feature map of channel c1.

[0337]

[0338] It should be noted that when the feature point to be encoded is feature point D2 within channel c2, a preset weight matrix corresponding to channel c2 can be determined. Next, to determine at least one first probability distribution parameter corresponding to feature point D2, linear weighting is performed on the feature value of the encoded feature point within the first target region in the feature map corresponding to the target channel group, and on the first feature of the feature point corresponding to the first target location within the second target region in the first feature matrix corresponding to the target channel group, based on the preset weight matrix corresponding to channel c2. The preset weight matrix corresponding to channel c2 is different from the preset weight matrix corresponding to channel c1.

[0339] It should be noted that regarding the method of determining at least one first probability distribution parameter corresponding to an encoding target feature point D1 when an encoding target location L1 within a second target region Q21 is determined as a first target location for channel c1 and a location L2 within a second target region Q22 is determined as a first target location for channel c2, reference is made to the description in the embodiment of FIG. 12b based on FIG. 12d. Details are not described again in this specification. Regarding the method of determining at least one first probability distribution parameter corresponding to an encoding target feature point D1 when a portion of a first unencoded location within a second target region Q21 and an encoding target location L1 are determined as a first target location for channel c1, and a location L2 within a second target region Q22 and a portion of a second unencoded location are determined as a first target location for channel c2, reference is made to the description in the embodiment of FIG. 12c based on FIG. 12d. Details are not described again in this specification.

[0340] It should be noted that when the number of channels k included in the target channel group exceeds 2, at least one first probability distribution parameter corresponding to the encoding target feature point can alternatively be determined by referring to the embodiment in FIG. 12d. Details are not described further in this specification.

[0341] FIG. 13a is a diagram of an example of a process for determining probability distribution parameters. In the embodiment of FIG. 13a, the number of channels k included in the target channel group is 1, ks1 = ks3 = 3, and the encoding order on the encoder side is illustrated in (2) of FIG. 3c (the encoder side determines the corresponding probability distribution parameter based on the estimated information matrix in the process of encoding feature points corresponding to the black blocks, and determines the corresponding probability distribution parameter based on S204 of this application in the process of encoding feature points corresponding to the white blocks). The target channel group is assumed to include a first channel (hereinafter referred to as channel c1). In the feature map of channel c1, the blocks filled with diagonal lines represent feature points D to be encoded, and feature points D to be encoded are feature points within the white checkerboard. In the first feature matrix of channel c1, the blocks filled with diagonal lines represent positions L to be encoded. Position L to be encoded is the position of feature point D to be encoded.

[0342] Refer to FIG. 13a. For example, a first target region Q1, located within the feature map of channel c1 and centered on the encoding target feature point D, can be determined based on the size of a linear weighting window, and a second target region Q2, located within the first feature matrix of channel c1 and centered on the encoding target position L, is determined based on the size of a linear weighting window. The sizes of the first target region Q1 and the second target region Q2 are the same, and both are ks1*ks2.

[0343] Refer to FIG. 13a. For example, the encoded feature points within the first target region Q1 can be determined as feature points for linear weighting calculation within the feature map, i.e., feature points corresponding to the black blocks within the first target region Q1 in FIG. 13a, and feature points at the upper left and upper right of the encoding target feature point D within the first target region Q1. The location of the encoding target feature point D is assumed to be (c1, w, h). In this case, the locations of the feature points for linear weighting calculation within the feature map of channel c1 include (c1, w-1, h-1), (c1, w, h-1), (c1, w+1, h-1), (c1, w-1, h), (c1, w, h+1), and (c1, w+1, h).

[0344] For example, a first target location for linear weighting calculation within a first feature matrix of channel c1 may be determined based on a location other than an encoded location within a second target region Q2 within the first feature matrix of channel c1. Refer to FIG. 13a. A location other than an encoded location within the second target region Q2 within the first feature matrix of channel c1 may be determined as a first target location, that is, a block filled with diagonal lines and a block filled with dots within the second target region Q2 in FIG. 13a. The location of the encoding target feature point D is assumed to be (c1, w, h), and the location of the feature point for linear weighting calculation within the feature map of channel c1 is assumed to include (c1, w-1, h-1), (c1, w, h-1), (c1, w+1, h-1), (c1, w-1, h), (c1, w, h+1), and (c1, w+1, h). In this case, the first target positions include (c1, w, h), (c1, w-1, h+1), and (c1, w+1, h+1).

[0345] For example, a preset weight matrix corresponding to channel c1 can be determined. The preset weight matrix includes a weight map. The size of the weight map is the same as the size of the linear weighting window, and the weight map includes weights of ks1*ks2 feature points as shown in FIG. 13a. The size of the weight map is 3*3, and the weight map includes 9 weights: ω1, ω2, ω3, ω4, ω5, ω6, ω7, ω8, and ω9.

[0346] For example, to obtain at least one first probability distribution parameter corresponding to an encoded target feature point D within a feature map of channel c1, linear weighting may be performed based on a weight map corresponding to channel c1, on the feature value of an encoded feature point within a first target region Q1 within the feature map of channel c1, and on a first feature of a feature point corresponding to a first target position within a second target region Q2 within the first feature matrix of channel c1.

[0347] The feature values ​​of the encoded feature points within the first target region Q1 in the feature map of channel c1 are represented as y[c1][w-1][h-1], y[c1][w][h-1], y[c1][w+1][h-1], y[c1][w][h+1], and y[c1][w+1][h], and the first features of the feature points corresponding to the first target positions within the second target region Q2 in the first feature matrix of channel c1 are and It is assumed to be represented as. In this case, at least one first probability distribution parameter = corresponding to the encoding target feature point D in the feature map of channel c1.

[0348]

[0349] It should be noted that regarding the method of determining at least one first probability distribution parameter corresponding to an encoding target feature point D in the feature map of channel c1 when an encoding target location L in the second target region Q2 in the first feature matrix of channel c1 is determined as the first target location, reference is made to the description in the embodiment of FIG. 12b based on the embodiment of FIG. 13a. Details are not described again in this specification. Regarding the method of determining at least one first probability distribution parameter corresponding to an encoding target feature point D in the feature map of channel c1 when a portion of the first unencoded location and an encoding target location L in the second target region Q2 in the first feature matrix of channel c1 are determined as the first target location, reference is made to the description in the embodiment of FIG. 12c based on the embodiment of FIG. 13a. Details are not described again in this specification.

[0350] It should be noted that when the number of channels k included in the target channel group exceeds 1, at least one first probability distribution parameter corresponding to at least one encoding target feature point D can be determined by referring to the description in the embodiment of FIG. 12d based on FIG. 13a. Details are not described further in this specification.

[0351] FIG. 13b is a diagram of an example of a process for determining probability distribution parameters. In the embodiment of FIG. 13b, the number of channels k included in the target channel group is 1, ks1 = ks3 = 3, and the encoding order on the encoder side is illustrated in (2) of FIG. 3c (the encoder side determines the corresponding probability distribution parameters based on S204 of this application in both the process of encoding feature points corresponding to black blocks and the process of encoding feature points corresponding to white blocks). It is assumed that the channels included in the target channel group include a first channel (hereinafter referred to as channel c1). In the feature map of channel c1, blocks filled with diagonal lines represent feature points D to be encoded, and feature points D to be encoded are feature points within the black checkerboard. In the first feature matrix of channel c1, blocks filled with diagonal lines represent positions L to be encoded. Position L to be encoded is the position of feature point D to be encoded.

[0352] Refer to FIG. 13b. For example, a first target region Q1, located within the feature map of channel c1 and centered on the encoding target feature point D, can be determined based on the size of a linear weighting window, and a second target region Q2, located within the first feature matrix of channel c1 and centered on the encoding target position L, is determined based on the size of a linear weighting window. The sizes of the first target region Q1 and the second target region Q2 are the same, and both are ks1*ks2.

[0353] Refer to FIG. 13b. For example, an encoded feature point within a first target region Q1 can be determined as a feature point for linear weighting calculation within a feature map, that is, a feature point corresponding to a black block within the first target region Q1 in FIG. 13b. The location of the encoding target feature point D is assumed to be (c1, w, h). In this case, the location of the feature point for linear weighting calculation within the feature map of channel c1 includes (c1, w-1, h-1) and (c1, w+1, h+1).

[0354] For example, a first target location for linear weighting calculation within the first feature matrix of channel c1 may be determined based on a location other than an encoded location within the second target region Q2. The encoded location is the location of the encoded feature point.

[0355] Refer to FIG. 13b. Locations other than the encoded locations within the second target region Q2 in the first feature matrix of channel c1 can be determined as the first target locations, i.e., the blocks filled with diagonal lines and the blocks filled with dots, and as locations within the second target region Q2 in FIG. 13b. The location of the encoding target feature point D is (c1, w, h), and the location of the feature point for linear weighting calculation within the feature map of channel c1 is assumed to include (c1, w-1, h-1) and (c1, w+1, h-1). In this case, the first target locations include (c1, w, h), (c1, w, h-1), (c1, w-1, h), (c1, w+1, h), (c1, w, h+1), (c1, w-1, h+1), and (c1, w+1, h+1).

[0356] For example, a preset weight matrix corresponding to channel c1 can be determined. The preset weight matrix includes a weight map. The size of the weight map is the same as the size of the linear weighting window, and the weight map includes weights of ks1*ks2 feature points as shown in FIG. 13b. The size of the weight map is 3*3, and the weight map includes 9 weights: ω1, ω2, ω3, ω4, ω5, ω6, ω7, ω8, and ω9.

[0357] For example, to obtain at least one first probability distribution parameter corresponding to an encoded target feature point D within a feature map of channel c1, linear weighting may be performed based on a weight map corresponding to channel c1, on the feature value of an encoded feature point within a first target region Q1 within the feature map of channel c1, and on a first feature of a feature point corresponding to a first target position within a second target region Q2 within the first feature matrix of channel c1.

[0358] The feature values ​​of the encoded feature points within the first target region Q1 in the feature map of channel c1 are represented as y[c1][w-1][h-1] and y[c1][w+1][h-1], and the first features of the feature points corresponding to the first target positions within the second target region Q2 in the first feature matrix of channel c1 are and It is assumed to be represented as. In this case, at least one first probability distribution parameter = corresponding to the encoding target feature point D in the feature map of channel c1.

[0359]

[0360] It should be noted that regarding the method of determining at least one first probability distribution parameter corresponding to an encoding target feature point D in the feature map of channel c1 when an encoding target location L in the second target region Q2 in the first feature matrix of channel c1 is determined as the first target location, reference is made to the description in the embodiment of FIG. 12b based on the embodiment of FIG. 13b. Details are not described again in this specification. Regarding the method of determining at least one first probability distribution parameter corresponding to an encoding target feature point D in the feature map of channel c1 when a portion of the first unencoded location and an encoding target location L in the second target region Q2 in the first feature matrix of channel c1 are determined as the first target location, reference is made to the description in the embodiment of FIG. 12c based on the embodiment of FIG. 13b. Details are not described again in this specification.

[0361] It should be noted that when the number of channels k included in the target channel group exceeds 1, at least one first probability distribution parameter corresponding to the encoding target feature point D can be determined by referring to the description in the embodiment of FIG. 12d based on FIG. 13b. Details are not described further in this specification.

[0362] For example, the following embodiments in FIGS. 14a and 14b describe a process in which linear weighting is performed on at least one feature value of at least one encoded feature point corresponding to a target channel group comprising k channels, and on a second feature matrix corresponding to a target channel group, in order to determine at least one second probability distribution parameter corresponding to an encoded target feature point corresponding to a target channel group.

[0363] FIG. 14a is a diagram of an example of a process for determining a probability distribution parameter. The embodiment in FIG. 14a describes a process for determining a second probability distribution parameter when a second feature matrix output by a hyperdecoder network as a second feature matrix of C channels includes a first feature matrix of C channels and a second feature matrix of C channels.

[0364] In the embodiment of FIG. 14a, the number of channels k included in the target channel group is 1, ks1 = ks3 = 3, and the encoding order on the encoder side is shown in (1) of FIG. 3c. The target channel group is assumed to include a first channel (hereinafter referred to as channel c1). In the feature map of channel c1, gray blocks represent encoded feature points, white blocks represent unencoded feature points, and blocks filled with diagonal lines represent feature points D to be encoded. In the second feature matrix of channel c1, gray blocks represent encoded locations, white blocks represent unencoded locations, and blocks filled with diagonal lines represent locations L to be encoded. The encoded location is the location of the encoded feature point, the unencoded location is the location of the unencoded feature point, and the encoding target location L is the location of the encoding target feature point D.

[0365] Refer to FIG. 14a. For example, a first target region Q1, located within the feature map of channel c1 and centered on the encoding target feature point D, can be determined based on the size of a linear weighting window, and a third target region Q3, located within the second feature matrix of channel c1 and centered on the encoding target position L, is determined based on the size of a linear weighting window. The sizes of the first target region Q1 and the third target region Q3 are the same, and both are ks1*ks2.

[0366] Refer to FIG. 14a. For example, the encoded feature points within the first target region Q1 can be determined as feature points for linear weighting calculation within the feature map, that is, feature points corresponding to the gray blocks within the first target region Q1 in FIG. 14a. The location of the encoding target feature point D is assumed to be (c1, w, h). In this case, the locations of the feature points for linear weighting calculation within the feature map of channel c1 include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1).

[0367] For example, a second target location for linear weighting calculation within the second feature matrix of channel c1 may be determined based on a location other than an encoded location within the third target region Q3 within the second feature matrix of channel c1. In a possible embodiment, an encoded target location L and at least one first unencoded location within the third target region Q3 within the second feature matrix of channel c1 may be determined as the second target location. The first unencoded location is a location other than the encoded target location L among all unencoded locations within the third target region Q3 within the second feature matrix of channel c1.

[0368] Refer to FIG. 14a. In a possible embodiment, locations other than the encoded locations within the third target region Q3 in the second feature matrix of channel c1 (i.e., the encoded target locations L and all first unencoded locations within the third target region Q3 in the second feature matrix of channel c1) may be determined as locations within the third target region Q3 in FIG. 14a that correspond to the second target locations, i.e., blocks filled with diagonal lines and blocks filled with dots.

[0369] It is assumed that the location of the encoding target feature point D is (c1, w, h), and the locations of the feature points for linear weighting calculation within the feature map of channel c1 include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1). In this case, the second target locations include (c1, w, h), (c1, w+1, h), (c1, w-1, h+1), (c1, w, h+1), and (c1, w+1, h+1).

[0370] For example, at least one difference corresponding to at least one encoded feature point in a first target region Q1 within a feature map of channel c1 can first be determined based on at least one feature value of at least one encoded feature point in a first target region Q1 within a feature map of channel c1, and at least one corresponding first probability distribution parameter, as illustrated in the block filled with vertical stripes in FIG. 14a (at least one first probability distribution parameter corresponding to at least one encoded feature point can first be determined in the manner described in the embodiments of FIG. 12a to 12d, and in the embodiments of FIG. 13a and 13b).

[0371] It is assumed that the location of the encoding target feature point D is (c1, w, h), the feature value of the encoded feature point in the first target region Q1 in the feature map of channel c1 is represented as y[c1][w-1][h], y[c1][w-1][h-1], y[c1][w][h-1], and y[c1][w+1][h-1], and the first probability distribution parameter corresponding to the encoded feature point in the first target region Q1 in the feature map of channel c1 is represented as m[c1][w-1][h], m[c1][w-1][h-1], m[c1][w][h-1], and m[c1][w+1][h-1]. The difference corresponding to the encoded feature point at position (c1, w-1, h) is represented by Diff[c1][w-1][h], the position corresponding to the encoded feature point at position (c1, w-1, h-1) is represented by Diff[c1][w-1][h-1], the difference corresponding to the encoded feature point at position (c1, w, h-1) is represented by Diff[c1][w][h-1], and the difference corresponding to the encoded feature point at position (c1, w+1, h-1) is represented by Diff[c1][w+1][h-1].

[0372] In a possible embodiment, the difference may be the difference between at least one feature value of at least one encoded feature point and at least one corresponding first probability distribution parameter.

[0373]

[0374] In a possible embodiment, the difference may be the absolute value of the difference between at least one feature value of at least one encoded feature point and at least one corresponding first probability distribution parameter. Diff[c1][w-1][h] is used as an example, i.e.,

[0375]

[0376] Here, abs represents obtaining the absolute value.

[0377] In a possible embodiment, the difference may be the square of the difference between at least one feature value of at least one encoded feature point and at least one corresponding first probability distribution parameter. Diff[c1][w-1][h] is used as an example, i.e.,

[0378]

[0379] For example, in the training process, the aggregation unit learns a weight matrix (i.e., a preset weight matrix) that corresponds to each channel and is used for linear weighting. The preset weight matrix may contain weight maps for k channels, and the size of the weight map is equal to the size of the linear weighting window, e.g., ks1*ks2. In the embodiment of FIG. 14a, the preset weight matrix for channel c1 contains a weight map for one channel, and the weight map corresponding to channel c1 has 9 weights: ω1 , ω2 , ω3 , ω4 , ω5 , ω6 , ω7 , 8 hours , and oh9 Includes

[0380] For example, to obtain at least one second probability distribution parameter corresponding to an encoded target feature point D within a feature map of channel c1, linear weighting may be performed based on a weight map corresponding to channel c1, on at least one difference corresponding to at least one encoded feature point within a first target region Q1 within the feature map of channel c1, and on a second feature of a feature point corresponding to a second target position within a third target region Q3 within a second feature matrix of channel c1.

[0381] Based on the above explanation, the second feature of the feature point corresponding to the second target position within the third target region Q3 in the second feature matrix of channel c1 is and It is assumed to be represented as. In this case, at least one second probability distribution parameter = corresponding to the encoding target feature point D in the feature map of channel c1.

[0382]

[0383] It should be noted that regarding the method of determining at least one second probability distribution parameter corresponding to an encoding target feature point D in the feature map of channel c1 when an encoding target location L in the third target region Q3 in the second feature matrix of channel c1 is determined as the second target location, reference is made to the description in the embodiment of FIG. 12b based on the embodiment of FIG. 14a. Details are not described again in this specification. Regarding the method of determining at least one second probability distribution parameter corresponding to an encoding target feature point D in the feature map of channel c1 when a portion of the first unencoded location and an encoding target location L in the third target region Q3 in the second feature matrix of channel c1 are determined as the second target location, reference is made to the description in the embodiment of FIG. 12c based on the embodiment of FIG. 14a. Details are not described again in this specification.

[0384] It should be noted that when the encoder side performs encoding in the order shown in (2) in FIG. 3c, at least one second probability distribution parameter corresponding to the encoding target feature point D in the feature map of channel c1 can be determined by referring to the description in FIG. 13a and FIG. 13b based on the embodiment in FIG. 14a. Details are not described again in this specification.

[0385] It should be noted that when the number of channels k included in the target channel group exceeds 1, at least one second probability distribution parameter corresponding to the encoding target feature point D can be determined by referring to the description in the embodiment of FIG. 12d based on FIG. 14a. Details are not described again in this specification.

[0386] FIG. 14b is a diagram of an example of a process for determining probability distribution parameters. The example in FIG. 14b describes a process for determining second probability distribution parameters when the estimated information matrix of C channels includes a first probability distribution parameter matrix of C channels and a second feature matrix of C channels.

[0387] In the embodiment of FIG. 14b, the number of channels k included in the target channel group is 1, ks1 = ks3 = 3, and the encoding order on the encoder side is shown in (1) of FIG. 3c. The target channel group is assumed to include a first channel (hereinafter referred to as channel c1). In the feature map of channel c1, gray blocks represent encoded feature points, white blocks represent unencoded feature points, and blocks filled with diagonal lines represent feature points D to be encoded.

[0388] In the embodiment of FIG. 14b, in the first probability distribution parameter matrix of channel c1 and the second feature matrix of channel c1, gray blocks represent encoded locations, white blocks represent unencoded locations, and blocks filled with diagonal lines represent encoding target locations L. Encoded locations are the locations of encoded feature points, unencoded locations are the locations of unencoded feature points, and encoding target locations L are the locations of encoding target feature points D.

[0389] Refer to FIG. 14b. For example, a first target region Q1, located within the feature map of channel c1 and centered on the encoding target feature point D, can be determined based on the size of a linear weighting window; a second target region Q2, located within the first probability distribution parameter matrix of channel c1 and centered on the encoding target location L, can be determined based on the size of a linear weighting window; and a third target region Q3, located within the second feature matrix of channel c1 and centered on the encoding target location L, can be determined based on the size of a linear weighting window. The sizes of the first target region Q1, the second target region Q2, and the third target region Q3 are the same and are all ks1*ks2.

[0390] Refer to FIG. 14b. For example, the encoded feature points within the first target region Q1 can be determined as feature points for linear weighting calculation within the feature map, that is, feature points corresponding to the gray blocks within the first target region Q1 in FIG. 14b. The location of the encoding target feature point D is assumed to be (c1, w, h). In this case, the locations of the feature points for linear weighting calculation within the feature map of channel c1 include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1).

[0391] Refer to FIG. 14b. For example, an encoded location within a second target region Q2, that is, a location corresponding to the gray block within the second target region Q2 in FIG. 14b, can be determined. It is assumed that the encoding target location L is (c1, w, h). In this case, the encoded locations within the second target region Q2 within the first probability distribution parameter matrix of channel c1 are (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1).

[0392] For example, as illustrated in the block filled with vertical stripes in FIG. 14b, at least one difference corresponding to at least one encoded feature point in a first target region Q1 within the feature map of channel c1 can first be determined based on at least one feature value of at least one encoded feature point in the first target region Q1, and at least one first probability distribution parameter of a feature point corresponding to an encoded position in a second target region Q2. For the method of determining at least one difference corresponding to at least one encoded feature point in the first target region Q1 within the feature map of channel c1, refer to the description in the embodiment in FIG. 14a. Details are not described again in this specification.

[0393] For example, a second target location for linear weighting calculation within the second feature matrix of channel c1 may be determined based on a location other than an encoded location within the third target region Q3 within the second feature matrix of channel c1. In a possible embodiment, an encoded target location L and at least one first unencoded location within the third target region Q3 within the second feature matrix of channel c1 may be determined as the second target location. The first unencoded location is a location other than the encoded target location L among all unencoded locations within the third target region Q3 within the second feature matrix of channel c1.

[0394] Refer to FIG. 14b. In a possible embodiment, locations other than the encoded locations within the third target region Q3 in the second feature matrix of channel c1 (i.e., the encoded target locations L and all first unencoded locations within the third target region Q3 in the second feature matrix of channel c1) may be determined as locations within the third target region Q3 in FIG. 14b that correspond to the second target locations, i.e., blocks filled with diagonal lines and blocks filled with dots.

[0395] It is assumed that the location of the encoding target feature point D is (c1, w, h), and the locations of the feature points for linear weighting calculation within the feature map of channel c1 include (c1, w-1, h), (c1, w-1, h-1), (c1, w, h-1), and (c1, w+1, h-1). In this case, the second target locations include (c1, w, h), (c1, w+1, h), (c1, w-1, h+1), (c1, w, h+1), and (c1, w+1, h+1).

[0396] For example, in the training process, the aggregation unit learns a weight matrix (i.e., a preset weight matrix) that corresponds to each channel and is used for linear weighting. The preset weight matrix may contain weight maps for k channels, and the size of the weight map is equal to the size of the linear weighting window, e.g., ks1*ks2. In the embodiment of FIG. 12a, the preset weight matrix for channel c1 contains a weight map for one channel, and the weight map corresponding to channel c1 has 9 weights: ω1 , ω2 , ω3 , ω4 , ω5 , ω6 , ω7 , 8 hours , and oh9 Includes

[0397] For example, to obtain at least one second probability distribution parameter corresponding to an encoded target feature point D within a feature map of channel c1, linear weighting may be performed based on a weight map corresponding to channel c1, on at least one difference corresponding to at least one encoded feature point within a first target region Q1 within the feature map of channel c1, and on a second feature of a feature point corresponding to a second target position within a third target region Q3 within a second feature matrix of channel c1.

[0398] Based on the above explanation, it is assumed that the difference corresponding to the encoded feature point at position (c1, w-1, h) is represented by Diff[c1][w-1][h], the position corresponding to the encoded feature point at position (c1, w-1, h-1) is represented by Diff[c1][w-1][h-1], the difference corresponding to the encoded feature point at position (c1, w, h-1) is represented by Diff[c1][w][h-1], and the difference corresponding to the encoded feature point at position (c1, w+1, h-1) is represented by Diff[c1][w+1][h-1]. The second feature of the feature point corresponding to the second target position within the third target region Q3 in the second feature matrix of channel c1 is and It is represented as. In this case, at least one second probability distribution parameter = corresponding to the encoding target feature point D in the feature map of channel c1.

[0399]

[0400] It should be noted that regarding the method of determining at least one second probability distribution parameter corresponding to an encoding target feature point D in the feature map of channel c1 when an encoding target location L in the third target region Q3 in the second feature matrix of channel c1 is determined as a second target location, reference is made to the description in the embodiment of FIG. 12b based on the embodiment of FIG. 14b. Details are not described again in this specification. Regarding the method of determining at least one second probability distribution parameter corresponding to an encoding target feature point D in the feature map of channel c1 when a portion of the first unencoded location and an encoding target location L in the third target region Q3 in the second feature matrix of channel c1 are determined as a second target location, reference is made to the description in the embodiment of FIG. 12c based on the embodiment of FIG. 14b. Details are not described again in this specification.

[0401] It should be noted that when the encoder side performs encoding in the order shown in (2) in FIG. 3, at least one second probability distribution parameter corresponding to the encoding target feature point D in the feature map of channel c1 can be determined by referring to the description in FIG. 13a and FIG. 13b based on the embodiment in FIG. 14b. Details are not described again in this specification.

[0402] It should be noted that when the number of channels k included in the target channel group exceeds 1, at least one second probability distribution parameter corresponding to the encoding target feature point D can be determined by referring to the description in the embodiment of FIG. 12d based on FIG. 14b. Details are not described further in this specification.

[0403] It should be understood that when the encoder side performs encoding in a different encoding order, the location of the encoded feature point in the feature map, the first target location in the first feature matrix, and the second target location in the second feature matrix may differ from the locations shown in the embodiments of FIGS. 12a to 12d, the embodiments of FIGS. 13a and 13b, and the embodiments of FIGS. 14a and 14b. The location of the encoded feature point in the feature map, the first target location in the first feature matrix, and the second target location in the second feature matrix are not limited in this application.

[0404] It should be noted that when the probability distribution model used by the probability estimation unit (V1) in the embodiment of FIG. 5 is a Gaussian distribution model having a mean of 0, that is, when the probability distribution parameter includes only the second probability distribution parameter (i.e., variance), the estimated information matrix output by the hyperdecoder network may include only the second feature matrix. In this case, the second probability distribution parameter (i.e., variance) may be determined in the manner described in the embodiments of FIG. 12a to 12d and FIG. 13a and FIG. 13b.

[0405] It should be understood that the method by which the decoder side determines at least one probability distribution parameter of a feature point to be decoded corresponds to the method by which the encoder side determines at least one probability distribution parameter of a feature point to be encoded. For details, refer to the description of the embodiment in FIGS. 12a through 12d, the description of the embodiment in FIGS. 13a and 13b, and the description of the embodiment in FIGS. 14a and 14b. Details are not described again in this specification.

[0406] For example, this application further provides a bitstream generation method to generate a bitstream according to the encoding method in the above-described embodiment.

[0407] For example, this application further provides a bitstream transmission method for transmitting a bitstream generated according to the bitstream generation method in the above-described embodiment.

[0408] For example, this application further provides a bitstream storage method to store a bitstream generated according to the bitstream generation method in the above-described embodiment.

[0409] In the example, FIG. 15 is a block diagram of a device (1500) according to an embodiment of this application. The device (1500) may include a processor (1501) and a transceiver / transceiver pin (1502), and optionally may further include a memory (1503).

[0410] The components of the device (1500) are connected together via a bus (1504). In addition to the data bus, the bus (1504) further includes a power bus, a control bus, and a status signal bus. However, for clarity of description, the various buses in the drawing are referred to as bus (1504).

[0411] Optionally, the memory (1503) may be configured to store instructions in the above-described method embodiment. The processor (1501) may be configured to execute instructions in the memory (1503), control a receiving pin to receive a signal, and control a transmitting pin to transmit a signal.

[0412] The device (1500) may be an electronic device or a chip of an electronic device in the above-described method embodiment.

[0413] All relevant details of the steps in the above-described method embodiments may be cited in the functional descriptions of the corresponding functional modules. Details are not described again in this specification.

[0414] The embodiment further provides a computer-readable storage medium. The computer-readable storage medium stores computer instructions. When computer instructions are operated on an electronic device, the electronic device may perform the associated method steps described above to implement the encoding and decoding methods of the embodiment described above.

[0415] The embodiments further provide a computer program product. When the computer program product is operated on a computer, the computer may perform the related steps described above to implement the encoding and decoding methods of the embodiments described above.

[0416] Additionally, embodiments of this application further provide a device. Specifically, the device may be a chip, a component, or a module. The device may include a processor and a memory connected to each other. The memory is configured to store computer-executable instructions. When the device is operated, the processor may execute computer-executable instructions stored in the memory, thereby causing the chip to perform the encoding and decoding methods of the method embodiments described above.

[0417] The electronic device, computer-readable storage medium, computer program product, or chip provided in the embodiments is configured to perform the corresponding method provided above. Therefore, regarding the beneficial effects that can be achieved, refer to the beneficial effects of the corresponding method provided above. Details are not described again in this specification.

[0418] Based on the description of the above-described embodiment, a person skilled in the art will understand that the division into the above-described functional modules is used as an example for the purpose of convenient and brief explanation. In actual applications, the above-described functions may be assigned to different functional modules to be implemented based on requirements. In other words, the internal structure of the device is divided into different functional modules to implement all or part of the functions described above.

[0419] It should be understood that in some of the embodiments provided in this application, the disclosed devices and methods may be implemented in different ways. For example, the described device embodiments are merely examples. For example, a division into modules or units is merely a logical functional division, and may be a different division during actual implementation. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not performed. Additionally, the mutual coupling or direct coupling or communication connection shown or discussed may be implemented through some interfaces. Indirect coupling or communication connection between devices or units may be implemented in an electronic form, a mechanical form, or another form.

[0420] Units described as separate parts may or may not be physically separate, and parts displayed as units may be one or more physical units, located in one place, or distributed across different places. Some or all of the units may be selected based on practical requirements to achieve the purpose of the solution of the embodiments.

[0421] Additionally, the functional units in the embodiments of this application may be integrated into a single processing unit, each unit may exist physically independently, or two or more units may be integrated into a single unit. The integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0422] Any of the embodiments of this application and any of the same embodiments may be freely combined. Any combination of the above contents will fall within the scope of this application.

[0423] When an integrated unit is implemented in the form of a software functional unit and is sold or used as an independent product, the integrated unit may be stored within a readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, or a part contributing to the prior art, or all or part of the technical solution may be implemented in the form of a software product. The software product is stored within a storage medium and includes some instructions for instructing a device or processor (which may be a single-chip microcomputer, a chip, etc.) to perform all or part of the steps of the method described in the embodiments of this application. The said storage medium includes any medium capable of storing program code, such as a USB flash drive, a removable hard disk drive, read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk.

[0424] The foregoing describes embodiments of this application with reference to the attached drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely examples, but are not limiting. A person skilled in the art who is inspired by this application may make many additional modifications without departing from the purpose of this application and the scope of protection of the claims, and all such modifications will be within the scope of this application.

[0425] The method or algorithm steps described in combination with the contents disclosed in the embodiments of this application may be implemented by hardware or by a processor by executing software instructions. Software instructions may include corresponding software modules. Software modules may be stored in Random Access Memory (RAM), flash memory, Read Only Memory (ROM), Erasable Programmable ROM (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), registers, hard disk drives, removable hard disk drives, compact disc read-only memory (CD-ROM), or any other form of storage medium widely known in the art. For example, the storage medium is coupled to the processor so that the processor can read information from the storage medium and write information into the storage medium. Certainly, the storage medium may be a component of the processor. The processor and the storage medium may be placed within an ASIC.

[0426] A person skilled in the art should recognize that in one or more of the examples above, the function described in the embodiments of this application may be implemented by hardware, software, firmware, or any combination thereof. When the function is implemented by software, the function may be stored in a computer-readable medium or transmitted as one or more instructions or codes in a computer-readable medium. A computer-readable medium includes a computer-readable storage medium and a communication medium, wherein the communication medium includes any medium that enables a computer program to be transmitted from one place to another. The storage medium may be any available medium accessible by a general-purpose or dedicated computer.

[0427] The foregoing describes embodiments of this application with reference to the attached drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely examples, but are not limiting. A person skilled in the art who is inspired by this application may make many additional modifications without departing from the purpose of this application and the scope of protection of the claims, and all such modifications will be within the scope of this application.

Claims

Claim 1 As an encoding method, the method comprises: acquiring an image to be encoded; generating a feature map of C channels based on the image to be encoded, wherein the feature map includes feature values ​​of a plurality of feature points, and C is a positive integer; generating an estimated information matrix of C channels based on the feature map of C channels; grouping the C channels into N channel groups, wherein N is an integer greater than 1, and each channel group includes k channels, and for each of the N channel groups, k is the same or different, and k is a positive integer; for at least one target channel group among the N channel groups, determining at least one probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group based on at least one feature value of at least one encoded feature point corresponding to the target channel group and an estimated information matrix corresponding to the target channel group; and determining a probability distribution corresponding to the encoding target feature point based on the at least one probability distribution parameter corresponding to the encoding target feature point. An encoding method comprising the step of encoding the encoding target feature point based on the probability distribution corresponding to the encoding target feature point. Claim 2 In claim 1, the step of determining at least one probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group based on at least one feature value of at least one encoded feature point corresponding to the target channel group and an estimated information matrix corresponding to the target channel group comprises the step of performing linear weighting on the at least one feature value of the at least one encoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group in order to determine the at least one probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group. Claim 3 In claim 2, the estimated information matrix comprises estimated information of a plurality of feature points, and in order to determine the at least one probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group, the step of performing linear weighting on the at least one feature value of the at least one encoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group comprises: a step of determining, based on the encoding target feature point, a first target region within a feature map corresponding to the target channel group and a second target region within the estimated information matrix corresponding to the target channel group; and a step of performing linear weighting on the at least one feature value of the at least one encoded feature point within the first target region and the estimated information of the at least one feature point within the second target region in order to obtain the at least one probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group. Claim 4 In claim 3, when k exceeds 1, the step of determining a first target region within a feature map corresponding to the target channel group and a second target region within the estimated information matrix corresponding to the target channel group based on the encoding target feature point comprises: a step of determining as the first target region an area of ​​a preset size centered on the encoding target feature point within the feature map of the first channel and a preset size area centered on a feature point corresponding to the location of the encoding target feature point within the feature map of the second channel - wherein the first channel corresponds to the encoding target feature point and the second channel is a channel other than the first channel within the target channel group -; and a step of determining as the second target region an area of ​​a preset size centered on the encoding target location within the estimated information matrix of the first channel and a preset size area centered on the location corresponding to the encoding target location within the estimated information matrix of the second channel - wherein the encoding target location is the location of the encoding target feature point - an encoding method comprising Claim 5 In claim 4, the step of performing linear weighting on at least one feature value of at least one encoded feature point within the first target region and the estimated information of at least one feature point within the second target region to obtain the at least one probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group comprises: a step of determining a first target location based on a location other than an encoded location within the second target region, wherein the encoded location is at least one location of the at least one encoded feature point; and a step of performing linear weighting on at least one feature value of at least one encoded feature point within the first target region and the estimated information of the feature point corresponding to the first target location to obtain the at least one probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group. Claim 6 In claim 5, when k exceeds 1, the step of determining a first target location based on a location other than an encoded location within the second target area comprises determining as the first target location a location corresponding to the encoded target location and at least one other unencoded location within the second target area within the estimated information matrix of the first channel, and the encoded target location and at least one other unencoded location within the second target area within the estimated information matrix of the second channel, wherein the first channel corresponds to the encoded target feature point and the second channel is a channel other than the first channel within the target channel group. Claim 7 In claim 5, when k exceeds 1, the step of determining a first target location based on a location other than an encoded location within the second target area comprises determining as the first target location a location corresponding to the encoding target location within the second target area within the estimated information matrix of the first channel, and the encoding target location within the second target area within the estimated information matrix of the second channel, wherein the first channel corresponds to the encoding target feature point, and the second channel is a channel other than the first channel within the target channel group. Claim 8 In claim 5, the step of performing linear weighting on the at least one feature value of the at least one encoded feature point within the first target region and the estimated information of the feature point corresponding to the first target location in order to obtain the at least one probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group comprises: the step of obtaining a pre-set weight matrix corresponding to the first channel corresponding to the encoding target feature point - the pre-set weight matrix includes a weight map of the k channels -; and the step of performing linear weighting on the at least one feature value of the at least one encoded feature point within the first target region and the estimated information of the feature point corresponding to the first target location based on the weight map of the k channels in order to obtain the at least one probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group. Claim 9 In claim 1, the estimated information matrix of the C channels comprises a first feature matrix of the C channels and a second feature matrix of the C channels, wherein the first feature matrix comprises a first feature of a plurality of feature points and the second feature matrix comprises a second feature of a plurality of feature points, and the probability distribution parameter comprises at least one first probability distribution parameter and at least one second probability distribution parameter, and the step of determining at least one probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group based on at least one feature value of an at least one encoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group comprises the step of determining at least one first probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group based on at least one feature value of the at least one encoded feature point corresponding to the target channel group and the first feature matrix corresponding to the target channel group; An encoding method comprising the step of determining at least one second probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group based on at least one feature value of the at least one encoded feature point corresponding to the target channel group and a second feature matrix corresponding to the target channel group. Claim 10 In claim 1, the estimated information matrix of the C channels comprises a first feature matrix of the C channels and a second probability distribution parameter matrix of the C channels, wherein the first feature matrix comprises a first feature of a plurality of feature points, the second probability distribution parameter matrix comprises a second probability distribution parameter of a plurality of feature points, and the at least one probability distribution parameter comprises at least one first probability distribution parameter, and the step of determining at least one probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group based on at least one feature value of at least one encoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group comprises the step of determining at least one first probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group based on at least one feature value of the at least one encoded feature point corresponding to the target channel group and the first feature matrix corresponding to the target channel group, and the step of determining a probability distribution corresponding to an encoding target feature point based on at least one probability distribution parameter corresponding to the encoding target feature point comprises a second probability distribution parameter corresponding to the target channel group An encoding method comprising: a step of determining a second probability distribution parameter corresponding to an encoding target feature point corresponding to a target channel group based on a matrix; and a step of determining the probability distribution corresponding to an encoding target feature point based on at least one first probability distribution parameter and at least one second probability distribution parameter corresponding to the encoding target feature point. Claim 11 In claim 1, the estimated information matrix of the C channels comprises a first probability distribution parameter matrix of the C channels and a second feature matrix of the C channels, wherein the first probability distribution parameter matrix comprises a first probability distribution parameter of a plurality of feature points, the second feature matrix comprises a second feature of a plurality of feature points, and the at least one probability distribution parameter comprises at least one second probability distribution parameter, and the step of determining at least one probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group based on at least one feature value of at least one encoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group comprises the step of determining at least one second probability distribution parameter corresponding to an encoding target feature point corresponding to the target channel group based on at least one feature value of the at least one encoded feature point corresponding to the target channel group and the second feature matrix corresponding to the target channel group, and the step of determining a probability distribution corresponding to an encoding target feature point based on at least one probability distribution parameter corresponding to the encoding target feature point comprises the first probability distribution parameter corresponding to the target channel group An encoding method comprising: a step of determining at least one first probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group based on a matrix; and a step of determining the probability distribution corresponding to the encoding target feature point based on the at least one first probability distribution parameter and the at least one second probability distribution parameter corresponding to the encoding target feature point. Claim 12 In claim 9, the step of determining at least one second probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group based on at least one feature value of at least one encoded feature point corresponding to the target channel group and a second feature matrix corresponding to the target channel group comprises: determining a first target region within a feature map corresponding to the target channel group and a third target region within the second feature matrix corresponding to the target channel group based on the encoding target feature point; determining at least one difference corresponding to the at least one encoded feature point within the first target region based on at least one feature value of at least one encoded feature point within the first target region and at least one corresponding first probability distribution parameter; and performing linear weighting on the second feature of at least one feature point within the third target region and the at least one difference corresponding to the at least one encoded feature point within the first target region in order to obtain the at least one second probability distribution parameter corresponding to the encoding target feature point corresponding to the target channel group. Claim 13 A bitstream generation method configured to generate a bitstream according to the encoding method according to any one of claims 1 to 12. Claim 14 A bitstream storage method configured to store a bitstream generated according to the bitstream generation method according to paragraph 13. Claim 15 A bitstream transmission method configured to transmit a bitstream generated according to the bitstream generation method according to paragraph 13. Claim 16 A decoding method comprising: receiving a bitstream; decoding the bitstream to obtain an estimated information matrix of C channels, wherein C is a positive integer; decoding the bitstream to obtain feature values ​​of feature points of C channels based on the estimated information matrix of C channels to obtain a feature map of C channels; determining, for a feature point to be decoded, a target channel group to which the channel corresponding to the feature point to be decoded belongs from N channel groups obtained by grouping the C channels, wherein each channel group includes k channels, and for each of the N channel groups, k is the same or different, k is a positive integer, and N is an integer greater than 1; and determining at least one probability distribution parameter corresponding to the feature point to be decoded based on at least one feature value of at least one decoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group. A decoding method comprising: a step of determining a probability distribution corresponding to a feature point to be decoded based on at least one probability distribution parameter corresponding to a feature point to be decoded based on the at least one probability distribution parameter corresponding to the feature point to be decoded based on the probability distribution corresponding to the feature point to be decoded based on the probability distribution corresponding to the feature point to be decoded based on the method for obtaining a feature value; and a step of performing reconstruction based on the feature map of C channels to output a reconstructed image. Claim 17 In claim 16, the step of determining at least one probability distribution parameter corresponding to the decoding target feature point based on at least one feature value of at least one decoded feature point corresponding to the target channel group and an estimated information matrix corresponding to the target channel group comprises the step of performing linear weighting on the at least one feature value of the at least one decoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group in order to determine the at least one probability distribution parameter corresponding to the decoding target feature point. Claim 18 A decoding method according to claim 17, wherein the estimated information matrix comprises estimated information of a plurality of feature points, and the step of performing linear weighting on the at least one feature value of the at least one decoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group in order to determine the at least one probability distribution parameter corresponding to the decoding target feature point comprises: a step of determining, based on the decoding target feature point, a first target region within a feature map corresponding to the target channel group and a second target region within the estimated information matrix corresponding to the target channel group; and a step of performing linear weighting on the at least one feature value of the at least one decoded feature point within the first target region and the estimated information of the at least one feature point within the second target region in order to obtain the at least one probability distribution parameter corresponding to the decoding target feature point. Claim 19 In claim 18, when k exceeds 1, the step of determining a first target region within a feature map corresponding to the target channel group and a second target region within the estimated information matrix corresponding to the target channel group based on the decoding target feature point comprises: determining as the first target region an area of ​​a preset size centered on the decoding target feature point within the feature map of the first channel and an area of ​​a preset size centered on a feature point corresponding to the location of the decoding target feature point within the feature map of the second channel - the first channel corresponds to the decoding target feature point, and the second channel is a channel other than the first channel within the target channel group -; and determining as the second target region an area of ​​a preset size centered on a decoding target location within the estimated information matrix of the first channel and an area of ​​a preset size centered on a location corresponding to the decoding target location within the estimated information matrix of the second channel - the decoding target location is the location of the decoding target feature point - a decoding method. Claim 20 A decoding method according to claim 19, wherein, in order to obtain the at least one probability distribution parameter corresponding to the at least one decoded feature point in the first target region, the step of performing linear weighting on the at least one feature value of the at least one decoded feature point in the first target region and the estimated information of the at least one feature point in the second target region comprises: the step of determining a first target location based on a location other than a decoded location in the second target region, wherein the decoded location is at least one location of the at least one decoded feature point; and the step of performing linear weighting on the at least one feature value of the at least one decoded feature point in the first target region and the estimated information of the feature point corresponding to the first target location in order to obtain the at least one probability distribution parameter corresponding to the at least one decoded feature point in the first target region. Claim 21 In claim 20, when k exceeds 1, the step of determining a first target location based on a location other than a decoded location within the second target region comprises determining as the first target location a location corresponding to the decoding target location and at least one other undecoded location within the second target region within the estimated information matrix of the first channel, and the decoding target location and at least one other undecoded location within the second target region within the estimated information matrix of the second channel, wherein the first channel corresponds to the decoding target feature point and the second channel is a channel other than the first channel within the target channel group. Claim 22 In claim 20, when k exceeds 1, the step of determining a first target location based on a location other than a decoded location within the second target region comprises determining as the first target location a location corresponding to the decoding target location within the second target region within the estimated information matrix of the first channel, and the decoding target location within the second target region within the estimated information matrix of the second channel, wherein the first channel corresponds to the decoding target feature point and the second channel is a channel other than the first channel within the target channel group. Claim 23 A decoding method according to claim 20, wherein, in order to obtain the at least one probability distribution parameter corresponding to the decoding target feature point, the step of performing linear weighting on the at least one feature value of the at least one decoded feature point within the first target region and the estimated information of the feature point corresponding to the first target location comprises: the step of obtaining a preset weight matrix corresponding to the first channel corresponding to the decoding target feature point - wherein the preset weight matrix includes a weight map of the k channels -; and the step of performing linear weighting on the at least one feature value of the at least one decoded feature point within the first target region and the estimated information of the feature point corresponding to the first target location based on the weight map of the k channels in order to obtain the at least one probability distribution parameter corresponding to the decoding target feature point. Claim 24 In claim 16, the estimated information matrix of the C channels comprises a first feature matrix of the C channels and a second feature matrix of the C channels, wherein the first feature matrix comprises a first feature of a plurality of feature points and the second feature matrix comprises a second feature of a plurality of feature points, and the at least one probability distribution parameter comprises at least one first probability distribution parameter and at least one second probability distribution parameter, and the step of determining at least one probability distribution parameter corresponding to the decoding target feature point based on at least one feature value of at least one decoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group comprises the step of determining at least one first probability distribution parameter corresponding to the decoding target feature point based on at least one feature value of at least one decoded feature point corresponding to the target channel group and the first feature matrix corresponding to the target channel group; A decoding method comprising the step of determining at least one second probability distribution parameter corresponding to a decoding target feature point based on at least one feature value of at least one decoded feature point corresponding to the target channel group and a second feature matrix corresponding to the target channel group. Claim 25 In claim 16, the estimated information matrix of the C channels comprises a first feature matrix of the C channels and a second probability distribution parameter matrix of the C channels, wherein the first feature matrix comprises a first feature of a plurality of feature points, the second probability distribution parameter matrix comprises a second probability distribution parameter of a plurality of feature points, and the at least one probability distribution parameter comprises at least one first probability distribution parameter, and the step of determining at least one probability distribution parameter corresponding to the decoding target feature point based on at least one feature value of at least one decoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group comprises the step of determining at least one first probability distribution parameter corresponding to the decoding target feature point based on at least one feature value of the at least one decoded feature point corresponding to the target channel group and the first feature matrix corresponding to the target channel group, and the step of determining a probability distribution corresponding to the decoding target feature point based on at least one probability distribution parameter corresponding to the decoding target feature point comprises a second probability distribution corresponding to the target channel group A decoding method comprising: a step of determining at least one second probability distribution parameter of the decoding target feature point based on a parameter matrix; and a step of determining the probability distribution corresponding to the decoding target feature point based on the at least one first probability distribution parameter and the at least one second probability distribution parameter corresponding to the decoding target feature point. Claim 26 In claim 16, the estimated information matrix of the C channels comprises a first probability distribution parameter matrix of the C channels and a second feature matrix of the C channels, wherein the first probability distribution parameter matrix comprises a first probability distribution parameter of a plurality of feature points, the second feature matrix comprises a second feature of a plurality of feature points, and the at least one probability distribution parameter comprises at least one second probability distribution parameter, and the step of determining at least one probability distribution parameter corresponding to the decoding target feature point based on at least one feature value of at least one decoded feature point corresponding to the target channel group and the estimated information matrix corresponding to the target channel group comprises the step of determining at least one second probability distribution parameter corresponding to the decoding target feature point based on at least one feature value of the at least one decoded feature point corresponding to the target channel group and the second feature matrix corresponding to the target channel group, and the step of determining a probability distribution corresponding to the decoding target feature point based on at least one probability distribution parameter corresponding to the decoding target feature point comprises the first probability distribution corresponding to the target channel group A decoding method comprising: a step of determining at least one first probability distribution parameter corresponding to the decoding target feature point based on a parameter matrix; and a step of determining the probability distribution corresponding to the decoding target feature point based on the at least one first probability distribution parameter and the at least one second probability distribution parameter corresponding to the decoding target feature point. Claim 27 A decoding method according to claim 24, wherein the step of determining at least one second probability distribution parameter corresponding to the decoding target feature point based on at least one feature value of at least one decoded feature point corresponding to the target channel group and a second feature matrix corresponding to the target channel group comprises: determining, based on the decoding target feature point, a first target region within a feature map corresponding to the target channel group and a third target region within the second feature matrix corresponding to the target channel group; determining at least one difference corresponding to at least one decoded feature point within the first target region based on at least one feature value of at least one decoded feature point within the first target region and at least one corresponding first probability distribution parameter; and performing linear weighting on the second feature of at least one feature point within the third target region and the at least one difference corresponding to at least one decoded feature point within the first target region in order to obtain the at least one second probability distribution parameter corresponding to the decoding target feature point. Claim 28 An encoder configured to perform the encoding method according to any one of claims 1 to 12. Claim 29 A decoder configured to perform the above decoding method according to any one of claims 16 to 27. Claim 30 An electronic device comprising a memory and a processor, wherein the memory is coupled to the processor, the memory stores a program instruction, and when the program instruction is executed by the processor, the electronic device is capable of performing a method according to any one of claims 1 to 12 and claims 16 to 27. Claim 31 A chip comprising one or more interface circuits and one or more processors, wherein the interface circuits are configured to receive a signal from a memory of an electronic device and transmit said signal to said processors, said signal includes computer instructions, and when said processors execute said computer instructions, the chip performs a method according to any one of claims 1 to 12 and claims 16 to 27. Claim 32 A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is operated on a computer or processor, the computer or processor can perform a method according to any one of claims 1 through 12 and claims 16 through 27. Claim 33 A computer program stored in a computer-readable storage medium, wherein the computer program comprises a software program, and when the software program is executed by a computer or processor, the computer or processor performs a method according to any one of claims 1 through 12 and claims 16 through 27.

Images