Method and apparatus for constructing merge list and deriving block matching model for compression of biomedical waveform signal

Abstract

A decoding method and apparatus, according to the present disclosure, comprise: generating prediction samples for a current subblock on the basis of a previously reconstructed reference subblock; generating residual samples for the current subblock on the basis of residual information for the current subblock obtained from a bitstream; and reconstructing the current subblock on the basis of the prediction samples and the residual samples, wherein the prediction samples may be generated on the basis of a first prediction mode based on a reference offset indicating a position of the reference subblock.

Description

Method and apparatus for constructing a merge list and deriving a block matching model for compression of biomedical waveform signals

[0001] The present invention relates to an apparatus and method for encoding a bitstream, a recording medium for storing an encoded bitstream, and an apparatus and method for decoding a stored or transmitted bitstream. More specifically, the invention relates to a method for dividing an input biomedical waveform signal into channel groups, frames, blocks, subblocks, etc., an apparatus and method for encoding the divided signals through prediction / conversion, a recording medium for storing a bitstream generated through encoding, and an apparatus and method for decoding a bitstream through inverse conversion / prediction.

[0002] Biomedical waveform compression is a technology used in the medical field to compress biological signals such as electrocardiograms (ECG), electroencephalograms (EEG), and electromyograms (EMG). It involves dividing each signal into multiple channels, subdividing each channel into a finite number of samples with timestamps at specific intervals, and then compressing these samples. In this process, the samples have discrete values.

[0003] Samples from biomedical waveform signals are segmented through a hierarchical structure, and the segmented samples can be compressed using hierarchical prediction and transformation units. Compression through prediction can be performed via intra-channel prediction, which utilizes temporal redundancy between previously reconstructed samples within a channel and the current sample, or via inter-channel prediction, which utilizes component-to-component redundancy between previously reconstructed samples within another channel and the current sample. Compression through transformation can be performed on the residual signals between the original signal and the predicted signal using transformations with low-frequency massing characteristics, such as DCT-based kernels, DST-based kernels, and identity kernels. In the case of lossy compression, the amount of data can be reduced by decreasing the amount of original information through frequency-based quantization, which removes high-frequency components. Finally, the predicted signal, residual signal, and parameters for reconstruction can reduce data through entropy encoding that utilizes statistical redundancy.

[0004] The compressed prediction signal, residual signal, and parameters for reconstruction can be restored to a biomedical waveform signal through entropy decoding, inverse quantization, inverse transform, and prediction. In the case of lossless compression, the restored biomedical waveform signal can be restored to be identical to the biomedical waveform signal before compression.

[0005] The present disclosure aims to provide a prediction method and apparatus based on a block matching mode.

[0006] The present disclosure aims to provide a method and apparatus for constructing a candidate list for a reference offset and determining a reference offset.

[0007] The present disclosure aims to provide a prediction method and apparatus utilizing a linear model in block matching mode.

[0008] The present disclosure aims to provide a prediction method and apparatus utilizing a weighted sum in block matching mode.

[0009] The technical problems to be solved in this disclosure are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which this disclosure belongs from the description below.

[0010] A decoding method according to the present disclosure comprises: generating prediction samples for a current subblock based on a previously restored reference subblock; generating residual samples for the current subblock based on residual information for the current subblock obtained from a bitstream; and restoring the current subblock based on the prediction samples and the residual samples, wherein the prediction samples may be generated based on a first prediction mode based on a reference offset indicating the location of the reference subblock.

[0011] In the decoding method according to the present disclosure, the reference offset is determined as a predicted reference offset, and when the reference offset is determined as a predicted reference offset, a candidate list for the reference offset may be formed.

[0012] In the decoding method according to the present disclosure, to construct the candidate list, at least one subblock may be searched among a previously restored subblock adjacent to the left of the current subblock, a subblock at the same location as the current subblock and restored in a previous channel, or a subblock at the left of the current subblock and restored in a previous channel.

[0013] In the decoding method according to the present disclosure, the candidate list may add as candidates the reference offset of a subblock generated in the first prediction mode among subblocks processed before the current subblock according to the coding order.

[0014] In the decoding method according to the present disclosure, the reference offset may be determined by determining a first reference offset based on index information of the candidate list obtained from the bitstream, and by performing a correction on the first reference offset to determine a second reference offset.

[0015] In the decoding method according to the present disclosure, the predicted samples are generated by applying at least one linear model parameter to the previously restored reference subblock, wherein the at least one linear model parameter can be derived to a value that minimizes the cost between the template region of the previously restored reference subblock and the second template region of the current subblock.

[0016] In the decoding method according to the present disclosure, at least one scaling parameter included in the at least one linear model parameter may be set to 1.

[0017] In the decoding method according to the present disclosure, the predicted samples are generated by applying at least one linear model parameter to the previously restored reference subblock, wherein the at least one linear model parameter may be derived based on a first average value for p(p≥1) samples having a maximum value among at least one sample in the template region of the reference subblock, a second average value for p(p≥1) samples having a maximum value among at least one sample in the template region of the current subblock, a third average value for p(p≥1) samples having a minimum value among at least one sample in the template region of the reference subblock, and a fourth average value for p(p≥1) samples having a minimum value among at least one sample in the template region of the current subblock.

[0018] In the decoding method according to the present disclosure, when the previously restored reference subblock includes at least two reference subblocks, the reference offset is determined for each reference subblock, and the reference offset for each reference subblock may be determined as a predicted reference offset.

[0019] In the decoding method according to the present disclosure, the predicted reference offset may be determined as either the reference offset of the last processed subblock among the subblocks generated in the first prediction mode that precedes the current subblock in coding order within the current channel, or the reference offset of the last in coding order subblock generated in the first prediction mode in the channel immediately preceding the current subblock.

[0020] In the decoding method according to the present disclosure, the predicted samples may be generated by calculating an average value for the difference between the template region of the previously restored reference subblock and the template region of the current subblock, and adding the average value to each sample of the previously restored reference subblock.

[0021] In a encoding method according to the present disclosure, the method comprises: generating prediction samples for a current subblock based on a previously restored reference subblock; generating residual samples for the current subblock based on the prediction samples for the current subblock; deriving transformation coefficients of the current subblock based on the residual samples; and encoding residual information regarding the transformation coefficients into a bitstream, wherein the prediction samples may be generated based on a first prediction mode based on a reference offset indicating the location of the reference subblock.

[0022] In a digital storage medium for storing a bitstream according to the present disclosure, the bitstream can be encoded by an encoding method.

[0023] According to the present disclosure, by providing a prediction method and apparatus based on a block matching mode, the prediction accuracy can be improved by precisely reflecting the repetitive pattern of a signal and the data to be encoded can be minimized to maximize data compression efficiency.

[0024] According to the present disclosure, signal compression performance can be improved by providing a method and apparatus for constructing a candidate list for a reference offset and determining a reference offset.

[0025] According to the present disclosure, by providing a prediction method and apparatus utilizing a linear model in block matching mode, adaptability to noise according to the signal measurement environment can be enhanced and data redundancy effectively eliminated, thereby improving encoding efficiency.

[0026] The effects obtainable from the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present disclosure pertains from the description below.

[0027] FIG. 1 is a diagram illustrating an example of generating a prediction signal through block matching according to one embodiment of the present disclosure.

[0028] FIG. 2 is a diagram illustrating an example of generating a prediction signal through block matching according to one embodiment of the present disclosure.

[0029] FIG. 3 is a diagram showing a biomedical waveform signal according to one embodiment of the present disclosure.

[0030] FIG. 4 is a diagram showing the hierarchical structure of a single biomedical waveform signal according to one embodiment of the present disclosure.

[0031] FIG. 5 is a diagram showing the hierarchical structure of a frame according to one embodiment of the present disclosure.

[0032] FIG. 6 is a flowchart of a biomedical signal segmentation process according to one embodiment of the present disclosure.

[0033] FIG. 7 is a diagram illustrating an example of generating a channel group for a biomedical waveform signal according to one embodiment of the present disclosure.

[0034] FIG. 8 is a flowchart of a biomedical waveform signal decoding method according to one embodiment of the present disclosure.

[0035] FIG. 9 is a diagram showing search positions for constructing a list of reference offsets according to one embodiment of the present disclosure.

[0036] FIG. 10 is a drawing showing an example of linear model derivation according to one embodiment of the present disclosure.

[0037] FIG. 11 is a drawing illustrating the derivation of a linear model based on a reference subblock and a current subblock according to one embodiment of the present disclosure.

[0038] FIG. 12 is a diagram showing the schematic configuration of a biomedical waveform signal decoding device according to one embodiment of the present disclosure.

[0039] FIG. 13 is a flowchart of a biomedical waveform signal encoding method according to one embodiment of the present disclosure.

[0040] FIG. 14 is a diagram showing the schematic configuration of a biomedical waveform signal encoding device according to one embodiment of the present disclosure.

[0041] A decoding method according to the present disclosure comprises: generating prediction samples for a current subblock based on a previously restored reference subblock; generating residual samples for the current subblock based on residual information for the current subblock obtained from a bitstream; and restoring the current subblock based on the prediction samples and the residual samples, wherein the prediction samples may be generated based on a first prediction mode based on a reference offset indicating the location of the reference subblock.

[0042] In the decoding method according to the present disclosure, the reference offset is determined as a predicted reference offset, and when the reference offset is determined as a predicted reference offset, a candidate list for the reference offset may be formed.

[0043] In the decoding method according to the present disclosure, to construct the candidate list, at least one subblock may be searched among a previously restored subblock adjacent to the left of the current subblock, a subblock at the same location as the current subblock and restored in a previous channel, or a subblock at the left of the current subblock and restored in a previous channel.

[0044] In the decoding method according to the present disclosure, the candidate list may add as candidates the reference offset of a subblock generated in the first prediction mode among subblocks processed before the current subblock according to the coding order.

[0045] In the decoding method according to the present disclosure, the reference offset may be determined by determining a first reference offset based on index information of the candidate list obtained from the bitstream, and by performing a correction on the first reference offset to determine a second reference offset.

[0046] In the decoding method according to the present disclosure, the predicted samples are generated by applying at least one linear model parameter to the previously restored reference subblock, wherein the at least one linear model parameter can be derived to a value that minimizes the cost between the template region of the previously restored reference subblock and the second template region of the current subblock.

[0047] In the decoding method according to the present disclosure, at least one scaling parameter included in the at least one linear model parameter may be set to 1.

[0048] In the decoding method according to the present disclosure, the predicted samples are generated by applying at least one linear model parameter to the previously restored reference subblock, wherein the at least one linear model parameter may be derived based on a first average value for p(p≥1) samples having a maximum value among at least one sample in the template region of the reference subblock, a second average value for p(p≥1) samples having a maximum value among at least one sample in the template region of the current subblock, a third average value for p(p≥1) samples having a minimum value among at least one sample in the template region of the reference subblock, and a fourth average value for p(p≥1) samples having a minimum value among at least one sample in the template region of the current subblock.

[0049] In the decoding method according to the present disclosure, when the previously restored reference subblock includes at least two reference subblocks, the reference offset is determined for each reference subblock, and the reference offset for each reference subblock may be determined as a predicted reference offset.

[0050] In the decoding method according to the present disclosure, the predicted reference offset may be determined as either the reference offset of the last processed subblock among the subblocks generated in the first prediction mode that precedes the current subblock in coding order within the current channel, or the reference offset of the last in coding order subblock generated in the first prediction mode in the channel immediately preceding the current subblock.

[0051] In the decoding method according to the present disclosure, the predicted samples may be generated by calculating an average value for the difference between the template region of the previously restored reference subblock and the template region of the current subblock, and adding the average value to each sample of the previously restored reference subblock.

[0052] In a encoding method according to the present disclosure, the method comprises: generating prediction samples for a current subblock based on a previously restored reference subblock; generating residual samples for the current subblock based on the prediction samples for the current subblock; deriving transformation coefficients of the current subblock based on the residual samples; and encoding residual information regarding the transformation coefficients into a bitstream, wherein the prediction samples may be generated based on a first prediction mode based on a reference offset indicating the location of the reference subblock.

[0053] In a digital storage medium for storing a bitstream according to the present disclosure, the bitstream can be encoded by an encoding method.

[0054] Embodiments of the present invention are described in detail with reference to the drawings attached to this specification so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0055] Throughout this specification, when a part is described as being 'connected' to another part, this includes not only cases where they are directly connected, but also cases where they are electrically connected with other elements in between.

[0056] Furthermore, throughout this specification, when a part is described as 'comprising' a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0057] Additionally, terms such as "first," "second," etc., may be used to describe various components, but said components should not be limited by said terms. These terms are used solely for the purpose of distinguishing one component from another.

[0058] Additionally, in the embodiments relating to the device and method described herein, some components of the device or some steps of the method may be omitted. Also, the order of some components of the device or some steps of the method may be changed. Additionally, other components or other steps may be inserted into some components of the device or some steps of the method.

[0059] In addition, some components or steps of the first embodiment of the present invention may be added to or replace some components or steps of the second embodiment of the present invention.

[0060] Furthermore, the components shown in the embodiments of the present invention are depicted independently to represent different characteristic functions and do not imply that each component consists of separate hardware or a single software unit. That is, for convenience of explanation, each component is described by listing it as a separate component, and at least two of the components may be combined to form a single component, or a single component may be divided into multiple components to perform a function. Such integrated and separated embodiments of each component are also included within the scope of the present invention as long as they do not deviate from the essence of the invention.

[0061] FIGS. 1 and FIGS. 2 are drawings illustrating an example of generating a prediction signal through block matching according to one embodiment of the present disclosure.

[0062] Specifically, FIG. 1 relates to the case where a prediction signal of the current subblock is generated from one reference subblock, and FIG. 2 relates to the case where a prediction signal of the current subblock is generated from two reference subblocks.

[0063] Block matching prediction technology is a technology that generates a prediction signal for the current subblock using a reference signal that has already been restored in the current channel. At this time, by signaling / parsing 1-bit flag information, it is possible to determine whether to generate a prediction signal for the current subblock from one reference subblock or from two reference subblocks. When generating a prediction signal for the current subblock from one reference subblock, the prediction signal can be calculated using the following Equation 1.

[0064]

[0065] Here, pred is the subblock signal currently to be predicted, rec is the signal already restored in the current channel, and t ref can represent an offset pointing to the location of the referenced subblock. Also, l k ≡ represents the length of the current subblock, and j represents the sample position of the current subblock.

[0066] When generating a prediction signal for the current subblock from two reference subblocks, the prediction signal can be calculated using the following mathematical formula 2.

[0067]

[0068] Here, pred is the subblock signal currently to be predicted, rec is the signal already restored in the current channel, and t ref,1 is the offset pointing to the location of the first referenced subblock, t ref,2 can represent an offset pointing to the location of the second referenced subblock. Also, l k ≡ represents the length of the current subblock, and j represents the sample position of the current subblock.

[0069] In Equation 1, the reference channel position offset t ref can be expressed as in mathematical equation 3. refClipping can be performed to be limited to a predefined range. In this case, the clipping range can be defined by an agreement between the decoder and the incoder, or determined by signaling / parsing a flag or index at a higher level.

[0070]

[0071] Here, t pred may represent offset information indicating the location of a subblock referenced by the most recently created subblock in block matching mode in the channel of the current subblock (hereinafter also referred to as the current channel), or offset information indicating the location of a subblock referenced by the most recently created subblock in block matching mode in the channel immediately preceding the current subblock. The most recently created subblock in block matching mode in the current channel may refer to the last processed subblock among the subblocks created in block matching mode that precede the current subblock in coding order in the current channel. Additionally, the most recently created subblock in block matching mode in the channel immediately preceding the current subblock may refer to the last in coding order subblock among the subblocks created in block matching mode before the processing time of the current subblock within the reference channel that is processed before the current channel.

[0072] Whether to use the location of the subblock referenced by the subblock most recently created in block matching mode in the current subblock's channel, or the location of the subblock referenced by the subblock most recently created in block matching mode in the channel immediately preceding the current subblock, can be determined by signaling / parsing a 1-bit flag. Additionally, the predicted reference offset t predIt can be adjusted based on the selected offset information, the size of the corresponding subblock, the size of the current subblock, the maximum allowable offset range, etc. Here, the selected offset information and the corresponding subblock may refer to either the subblock most recently created in block matching mode in the channel of the aforementioned current subblock or the subblock most recently created in block matching mode in the previous channel. As an example, if the size of the current subblock and the size of the corresponding subblock are different, the predicted reference offset may be scaled based on the size difference between the two subblocks. The maximum allowable offset range may be determined by signaling / parsing an index or flag at the upper unit.

[0073] Equation 3 can be extended to the case using two reference subblocks. In the case using two reference subblocks, t for each reference subblock pred It can signal / parse a 1-bit flag for each referenced subblock to determine whether to use the location of the subblock referenced by the subblock most recently created in block matching mode in the channel of the current subblock, or the location of the subblock referenced by the subblock most recently created in block matching mode in the channel immediately preceding the current subblock.

[0074] t in mathematical equation 3 diff can be determined through signaling / parsing, and t diff Signaling / parsing can be performed by separating the sign and absolute value of. t through signaling / parsing diff The determination of can be extended to the case using two reference subblocks, and t for each reference subblock diff Signaling / parsing can be performed by separating the sign and absolute value.

[0075] A final prediction signal can be generated by performing filtering on the current subblock generated from mathematical expression 1 or mathematical expression 2, and a 1-bit flag can be signaled / parsed to determine whether to perform filtering. In this case, the filter can be a smoothing filter or an interpolation filter.

[0076] A biomedical waveform signal may be a signal discretized and digitized in units of channels and timestamps to enable the waveform of a biological signal generated in the human body to be stored in a storage device or transmitted and received via a network device. A biomedical waveform signal may be a signal in which samples existing in a specific channel and a specific timestamp are collected for all channels and all timestamps. A channel group may be a single channel or multiple channels that group a biological signal composed of multiple channels to enable parallel processing. If two channels belong to the same channel group, they may be able to reference each other. If two channels belong to different channel groups, they may not be able to reference each other. A frame may be a set of samples corresponding to a single timestamp or multiple timestamps that group samples composed of multiple timestamps within a channel group to enable parallel processing. If there is no dependency between two frames, the reconstruction of the two frames may be parallelized. If there is a dependency between two frames, the reconstruction of the other frame may begin after the reconstruction of one frame is completely finished. A block may be a set of samples corresponding to a single timestamp or multiple timestamps within a frame. Channels belonging to the same block may all have the same number of samples. A subblock may be a set of samples per channel within a block. A block consisting of N channels may have N subblocks. Here, N may be an integer greater than or equal to 1. A sample may be a discretized and digitized value of a specific channel at a specific timestamp.

[0077] FIG. 3 is a diagram showing a biomedical waveform signal according to one embodiment of the present disclosure.

[0078] As illustrated in FIG. 3, multiple biomedical waveform signals may exist. The biomedical waveform signals may include various biomedical signals such as electrocardiograms, electroencephalograms, and electromyograms. The selected biomedical waveform signals may include signals of arbitrary length for each channel, and the number of channels may be arbitrarily set.

[0079] FIG. 4 is a diagram showing the hierarchical structure of a single biomedical waveform signal according to one embodiment of the present disclosure.

[0080] A selected single biomedical waveform signal consisting of multiple channels can be hierarchically partitioned in the order of channel groups and frames.

[0081] To split and select channel groups, some channels within a biomedical waveform signal can be grouped into a single channel group. In this case, the length of each channel belonging to the same channel group can be set to be equal. A channel group may consist of only a single channel or may consist of all channels within the biomedical waveform signal. Each channel group within a single biomedical waveform signal may consist of a different number of channels. Parameter information regarding how to split the channel groups can be determined by signaling / parsing at the biomedical waveform level.

[0082] To split and select frames, a channel group can be divided into frames of a specific length. A channel group can be divided into a single frame or multiple frames. In this case, the length of each channel belonging to the same frame can be set to be the same. A frame may consist of all channels within the channel group. If a channel group is divided into multiple frames, each frame may consist of different lengths or the same length. When restoring a specific frame, signaling / parsing can be performed on a frame-by-frame basis to determine whether it is an independent frame or a dependent frame, distinguishing between cases where the frame references no other frames and cases where the frame references one or more frames.

[0083] FIG. 5 is a diagram showing the hierarchical structure of a frame according to one embodiment of the present disclosure.

[0084] As illustrated in FIG. 5, hierarchical partitioning can be performed from the frame to the block and then to the sub-block. To partition and select the frame, the frame to be encoded / decoded can be selected. The frame to be encoded / decoded can be an independent frame or a dependent frame.

[0085] To divide and select blocks, a frame can be divided into blocks of a specific length. In this case, the length of each channel belonging to the same block can be set to be equal. A block may consist of only one sample or all samples within the frame.

[0086] To divide and select sub-blocks, a block composed of multiple channels can be divided into sub-blocks for each channel.

[0087] FIG. 6 is a flowchart of a biomedical signal segmentation process according to one embodiment of the present disclosure.

[0088] Biomedical waveform signals can be divided into multiple channels or channel groups consisting of a single channel. Each channel group can be divided into frames. Through signaling / parsing on a frame basis, blocks of a specific length can be divided from the frames. Sub-blocks can be divided from the blocks. The order of each division process can be changed or omitted.

[0089] Referring to FIG. 6, a biomedical waveform signal can be divided to generate multiple channel groups (S610). Each channel group may consist of multiple channels or a single channel.

[0090] According to one embodiment, when a channel group is composed of a plurality of channels, some or all of the channels within the channel group may be continuous or discontinuous.

[0091] FIG. 7 is a diagram illustrating an example of generating a channel group for a biomedical waveform signal according to one embodiment of the present disclosure.

[0092] Referring to Fig. 7, CG_0 and CG_2 illustrate the case where the channels within a channel group are discontinuous.

[0093] Referring to Fig. 7, CG_1 illustrates a case where the channels within a channel group are continuous.

[0094] According to one embodiment, information for generating a channel group from a biomedical waveform signal is defined in a set of waveform-unit parameters and can be signaled / parsed on a waveform-unit basis.

[0095] Referring to FIG. 6, the channel group can be divided into frames (S620).

[0096] According to one embodiment, a channel group can be divided into frames having a specific length. The channel group may be divided into a single frame or into multiple frames. If divided into multiple frames, each frame may be an independent frame or a dependent frame.

[0097] Independent frames can be restored regardless of previously restored frame information, while dependent frames can be restored only after all independent and dependent frames corresponding to the restoration order within the same channel group have been restored.

[0098] According to one embodiment, information regarding whether the current frame is an independent frame or a dependent frame can be signaled / parsed on a frame-by-frame basis.

[0099] According to one embodiment, encoding parameter information, such as the availability of specific prediction modes like inter-channel prediction and block matching, detailed prediction methods for each prediction mode, quantization levels, and maximum and minimum block sizes within a frame, can be signaled / parsed on a frame-by-frame basis so that any decoder can use it in the decoding process. Furthermore, these are merely examples, and encoding parameter information related to more prediction modes or transformation modes, etc., can be signaled / parsed on a frame-by-frame basis, and any decoder can use it in the decoding process.

[0100] A dependent frame can inherit encoding parameter information from an independent frame or a dependent frame that precedes it in the restoration order. When a dependent frame inherits encoding parameter information from a preceding independent or dependent frame, it may inherit all encoding parameters, or it may inherit only some parameter information and determine the remaining encoding parameters by newly signaling / parsing.

[0101] Referring to FIG. 6, blocks can be divided from a frame (S630).

[0102] According to one embodiment, the minimum and maximum lengths of a block may be defined in a frame unit parameter set, and block partitioning may be performed using such information.

[0103] According to one embodiment, the minimum and maximum lengths of a block may be omitted from signaling / parsing if any decoder can derive them on a frame-by-frame basis, or signaling / parsing if any decoder cannot derive them. Each block within a single frame may be composed of different lengths or the same length.

[0104] According to one embodiment, when restoring a specific block on a frame-by-frame basis, the independent block or dependent block can be determined by dividing the case where there is no other block within the same frame referenced by the block and the case where there is. For example, if the timestamp of the first sample of a specific block in any decoder is the same as the timestamp of the first sample of the frame, the block can be determined as an independent block. In this case, the encoder can omit signaling / parsing for the relevant information. For example, if the timestamp of the first sample of a specific block in any decoder within the same frame exists after the timestamp of the last sample of an independent block, the block can be determined as a dependent block. In this case, the encoder can omit signaling / parsing for the relevant information.

[0105] Referring to FIG. 6, the block can be divided into sub-blocks (S640).

[0106] According to one embodiment, the block can be divided into a plurality of sub-blocks according to the channel, and each sub-block can be composed of one channel.

[0107] According to one embodiment, whether it is an intra-channel prediction or an inter-channel prediction can be signaled / parsed at the sub-block level. Prediction mode information, parameter information for the prediction mode, residual samples, etc., can be signaled / parsed at the sub-block level.

[0108] FIG. 8 is a flowchart of a biomedical waveform signal decoding method according to one embodiment of the present disclosure.

[0109] Finally, restoration can be performed by conducting prediction and inverse transformation on the determined units, modes, and methods. The order of each process may be changed or omitted.

[0110] Referring to FIG. 8, prediction samples for the current subblock can be generated based on the previously restored reference subblock (S810).

[0111] A prediction unit can be determined. According to one embodiment, the prediction unit in the encoder / decoder may be a single subblock. Here, the subblock may be generated by dividing a biomedical waveform signal, as seen with reference to FIG. 6.

[0112] A prediction type (such as intra-channel prediction or cross-channel prediction) can be determined for each block. According to one embodiment, the prediction type of each prediction unit may be one of intra-channel prediction, cross-channel prediction, etc.

[0113] Among the prediction types determined for each prediction unit, it is possible to decide which mode to use to perform the prediction.

[0114] According to one embodiment, when in-channel prediction is performed for the current subblock, the prediction mode of the current subblock may be determined as a DC mode that generates the prediction signal of the current subblock as the average of the previously restored samples adjacent to the left of the current subblock.

[0115] According to one embodiment, when in-channel prediction is performed for a current subblock, the prediction mode of the current subblock may be determined as a pre-fit mode that derives a curve equation using the values ​​of previously restored samples adjacent to the current subblock and generates a prediction signal of the current subblock by extrapolating the sample values ​​of the current subblock from the derived curve.

[0116] According to one embodiment, when in-channel prediction is performed for the current subblock, the prediction mode of the current subblock may be a block matching mode that generates a prediction signal of the current subblock using a previously restored subblock as a reference block.

[0117] According to one embodiment, when cross-channel prediction is performed for a current subblock, it may be determined to be a cross-channel prediction mode that predicts the subblock of the current channel from a previously restored signal of a different channel corresponding to the location of the current subblock.

[0118] According to one embodiment, when the prediction mode of the current subblock is DC mode, the current subblock can be predicted by the average of the current subblock and the previously restored samples adjacent to the left. In the process of calculating the average, the number of previously restored samples adjacent to the left to be used may be predefined in the decoder or determined by signaling / parsing at the subblock or a higher level.

[0119] According to one embodiment, if the prediction mode of the current subblock is a pre-fit mode, the equation of a curve can be derived using the current subblock and the pre-recovered samples adjacent to the left, and the current subblock can be extrapolated to generate a prediction block. In the process of deriving the equation of the curve, the number of pre-recovered samples adjacent to the left to be used may be predefined in the decoder or determined by signaling / parsing at the subblock or a higher level.

[0120] According to one embodiment, when the prediction mode of the current subblock is a block matching mode, a prediction signal of the current subblock can be generated from a reference subblock. The reference subblock may refer to a block pointed to by a reference offset in the current subblock. For example, in the case of a block matching mode, a prediction signal of the current subblock can be generated through Equation 4.

[0121]

[0122] Here, t is the reference offset, rec is the signal already restored in the current channel, l k ≡ represents the length of the current subblock, and j represents the sample position of the current subblock.

[0123] According to one embodiment, when the prediction mode of the current subblock is a block matching mode, the reference offset may be composed of the sum of the predicted reference offset and the residual reference offset, and the predicted reference offset may be derived from the previously restored subblock and the residual reference offset may be signaled / parsed to finally determine the reference offset value. Additionally, clipping may be performed so that the finally determined reference offset is limited to a predefined range. At this time, the clipping range may be defined by an agreement between the decoding and decoding units, or determined by signaling / parsing a flag or index at the upper unit.

[0124] According to one embodiment, the predicted reference offset may be offset information of a subblock most recently created in block matching mode in the current channel or offset information of a subblock created in block matching mode in the channel immediately preceding the current subblock, and which offset value to use can be determined by signaling / parsing a 1-bit flag. The predicted reference offset may be adjusted based on the size of the subblock corresponding to the selected offset information, the size of the current subblock, the maximum allowable offset range, etc. Here, the subblock corresponding to the selected offset information may refer to either the subblock most recently created in block matching mode in the channel of the aforementioned current subblock or the subblock most recently created in block matching mode in the channel immediately preceding the current subblock. As an example, if the size of the current subblock and the size of the corresponding subblock are different, the predicted reference offset may be scaled based on the size difference between the two subblocks. The maximum allowable offset range may be determined by signaling / parsing an index or a flag at the upper unit.

[0125] According to one embodiment, when signaling / parsing a residual reference offset, the sign and absolute value of the residual reference offset can be separated for signaling / parsing.

[0126] According to one embodiment, if the prediction mode of the current subblock is a block matching mode, the final reference offset can be determined as the predicted reference offset.

[0127] According to one embodiment, in the process of determining a reference offset, whether to determine the predicted reference offset as the final reference offset, or to determine the residual channel offset by signaling / parsing and determining the final reference offset as the sum of the residual channel offset and the predicted reference offset, can be determined by signaling / parsing a 1-bit flag.

[0128] According to one embodiment, when determining the final reference offset with the predicted reference offset, a list of reference offsets can be constructed.

[0129] According to one embodiment, when constructing a list for a reference offset, at least one of previously restored subblocks adjacent to the left of the current subblock, previously restored subblocks at the same location as the current subblock and in the previous channel, and previously restored subblocks at the location to the left of the current subblock can be searched to construct the list.

[0130] FIG. 9 is a diagram showing search positions for constructing a list of reference offsets according to one embodiment of the present disclosure.

[0131] l k is the length of the current subblock, s m,k In this case, m is the channel index, k is the index indicating which block it is within the current frame, and A0, A1, B0, B1, etc. may be the locations of the blocks to be searched.

[0132] Referring to Fig. 9, blocks can be searched in the order A0->A1->B0->B1->B2->C0 and it can be searched whether the corresponding block is predicted in block matching mode.

[0133] However, the above-disclosed embodiment is merely one example, and the search location and search order may be changed.

[0134] According to one embodiment, when the prediction mode of the current subblock is a block matching mode, a reference offset is determined and a prediction signal is generated, and then the reference offset used in the process can be stored in a data storage space. For example, the data storage space may be a buffer. In addition, one or more reference offsets may be stored in the data storage space, and the number of data items to be stored may be predefined in the encoder / decoder.

[0135] According to one embodiment, when the prediction mode of the current subblock is block matching mode, during the process of constructing the reference offset list, the reference offset of the subblock most recently created in block matching mode can be added to the list. For example, a reference offset stored in the data storage space can be added to the reference offset list, and multiple or a single reference offset can be added to the reference offset list.

[0136] According to one embodiment, if a subblock generated in block matching mode is found during the search process, the reference offset used in that subblock can be stored in a reference offset list. The size of the reference offset list can be determined by pre-defining, block units, or signaling / parsing at a level higher than that unit. If the list of reference block locations becomes full during the search process, the search process may be terminated early.

[0137] According to one embodiment, if a number of reference subblocks are mixed during the search process and there is a subblock that has been restored in block matching mode, all reference offsets of the corresponding subblock can be added to the reference offset list.

[0138] According to one embodiment, if a reference offset list is configured, the final reference offset of the current subblock can be determined by signaling / parsing the index of the list.

[0139] According to one embodiment, when the reference offset of the current subblock is determined by signaling / parsing the index of the reference offset list, an additional correction value for the reference offset can be signaled / parsed to determine the corrected reference offset as the final reference offset, and whether to perform the correction can be determined by signaling / parsing a 1-bit flag.

[0140] According to one embodiment, when signaling / parsing additional correction values ​​for a reference offset, the sign and absolute value of the correction value can be separated and signaled / parsed. The absolute value of the correction value may be predefined in the form of a table in the encoder / decoder, and the absolute value of the correction value can be determined by signaling / parsing the index of the table. Additionally, the absolute value information may be signaled / parsed in units of waveform parameters to modify the value existing in the table. For example, the absolute value of the correction value may be defined in a table as in Equation 5, and the absolute value of the correction value can be determined by signaling the index of the table.

[0141]

[0142] Equation 5 is an example of the absolute value for the correction values ​​predefined in the table, and these values ​​may be modified or changed, and the size of the table may also be changed or modified.

[0143] According to one embodiment, when the prediction mode of the current subblock is a block matching mode, a linear model (or linear equation) can be applied to the reference subblock to generate a prediction signal for the current subblock. For example, a prediction signal for the current subblock can be generated as shown in Equation 6.

[0144]

[0145] Here, α block is the scaling parameter, β block can represent bias parameters.

[0146] According to one embodiment, linear model parameter α block and β block can be derived from the template area of ​​the reference subblock and the template area of ​​the current subblock, and the template area may refer to a set of N left samples adjacent to the reference subblock or the current subblock. N can be an integer greater than or equal to 1.

[0147] According to one embodiment, when N sample regions restored on the left are defined as a template, N can use a predefined value.

[0148] According to one embodiment, when N sample regions restored on the left are defined as a template, N can be determined based on the size of the current subblock.

[0149] According to one embodiment, when N sample regions restored on the left are defined as a template, N can be determined by signaling / parsing in block units, frame units, or channel group units.

[0150] According to one embodiment, when determining N by signaling / parsing, the value for N can be determined by directly signaling / parsing the value for N or by pre-defining a table for N values ​​and signaling / parsing an index for the table.

[0151] According to one embodiment, a linear model can be derived from the template region of a reference subblock and the template region of a current subblock, and linear model parameters, namely scaling parameters and bias parameters, that minimize the cost (e.g., Mean Square Error (MSE)) between the template region of the current subblock and the template region of the reference subblock can be derived. For example, the scaling parameter α through Equation 7 block and bias parameter β block It can induce.

[0152]

[0153] Here, rec[m] is the signal restored from the current channel m, t is the reference offset, and N is the size of the template area.

[0154] According to one embodiment, when deriving linear model parameters, 1-bit flag information can be signaled / parsed to set the scaling parameter to 1 and derive only the bias parameter. The simplified linear model derivation process can be represented as Equation 8.

[0155]

[0156] Here, rec[m] is the signal restored from the current channel m, t is the reference offset, and N is the size of the template area.

[0157] According to one embodiment, when deriving a linear model, correction of the scaling parameter may be performed. Whether correction is performed can be determined by signaling / parsing a 1-bit flag, and if the scaling parameter is corrected, the correction value can be determined by directly signaling / parsing it. Alternatively, the correction value may be defined in advance in the encoder / decoder, or it may be defined in the form of a table. If the correction value is defined in the form of a table, the correction value can be determined by signaling / parsing the index of the table.

[0158] According to one embodiment, after deriving linear model parameters, the derived linear model can be applied to a reference subblock to generate a prediction signal of the current subblock. For example, Equation 6 may represent a process of generating a prediction signal of the current subblock through the corresponding process.

[0159] According to one embodiment, when deriving linear model parameters, the linear model may be derived using only T samples at specific locations in the template region to simplify the derivation process. T is an integer greater than or equal to 1 and may be predefined in the encoder / decoder.

[0160] FIG. 10 is a drawing showing an example of linear model derivation according to one embodiment of the present disclosure.

[0161] Referring to Fig. 10, a linear model can be derived using only T samples. For example, T can be 4.

[0162] The location and / or number of specific samples used to derive the linear model can be determined based on the size of the current subblock.

[0163] According to one embodiment, when deriving linear model parameters, a linear model can be derived using the average value of the largest p samples in the template region of the reference subblock, the average value of p samples in the corresponding template region of the current subblock, the average value of the smallest p samples in the template region of the reference subblock, and the average value of p samples in the corresponding template region of the current subblock.

[0164] FIG. 11 is a drawing illustrating the derivation of a linear model based on a reference subblock and a current subblock according to one embodiment of the present disclosure.

[0165] Referring to FIG. 11, the average value (x) of the largest p (e.g., p=2) samples in the template region of the reference subblock refa) and the average value of the p largest samples in the corresponding template region of the current subblock (x cura ), the average value of the p smallest samples in the template region of the reference subblock (x refb The average value of the p smallest samples (x) in the template region of the current subblock corresponding to this. curb Linear model parameters α and β can be derived from ). Here, p can be an integer greater than or equal to 1 and may be predefined in the encoder / decoder.

[0166] The α and β derived from the corresponding process can be expressed as in Equation 9.

[0167]

[0168] According to one embodiment, when the prediction mode of the current subblock is a block matching mode, whether to generate the prediction subblock signal of the current subblock as a reference subblock or to generate it by applying a linear model to the reference subblock can be determined by signaling / parsing a 1-bit flag.

[0169] According to one embodiment, when the prediction mode of the current subblock is a block matching mode, a prediction signal of the current subblock can be generated using one reference subblock or multiple reference subblocks. Whether to use one reference subblock or multiple reference subblocks can be determined by signaling / parsing a 1-bit flag.

[0170] According to one embodiment, when the prediction mode of the current subblock is a block matching mode and multiple reference subblocks are used, the number of reference blocks u may be signaled / parsed to explicitly determine the number of reference blocks, or the number of reference blocks may not be signaled / parsed. If the number of reference blocks is not signaled / parsed, the number of reference blocks may be implicitly determined to be two. Here, u may be an integer greater than or equal to 2.

[0171] According to one embodiment, when the prediction mode of the current subblock is a block matching mode and multiple reference subblocks are used, the reference offset pointing to each reference subblock from the current subblock may be composed of the sum of the predicted reference offset and the residual reference offset, and the predicted reference offset may be derived for each reference offset and the residual reference offset may be signaled / parsed for each reference offset to finally determine the reference offset value for each subblock.

[0172] According to one embodiment, when the current subblock is in a block matching mode using multiple reference subblocks, each predicted reference offset may be offset information of the subblock most recently created in block matching mode in the current channel or offset information of the subblock created in block matching mode in the channel immediately preceding the current subblock, and which offset value to use can be determined by signaling / parsing a 1-bit flag for each predicted reference offset. The predicted reference offset may be adjusted based on the size of the subblock corresponding to the selected offset information, the size of the current subblock, the maximum allowable offset range, etc. Here, the subblock corresponding to the selected offset information may refer to either the subblock most recently created in block matching mode in the channel of the aforementioned current subblock or the subblock most recently created in block matching mode in the channel immediately preceding the current subblock. As an example, if the size of the current subblock and the size of the corresponding subblock are different, the predicted reference offset may be scaled based on the size difference between the two subblocks. The maximum allowable offset range may be determined by signaling / parsing an index or a flag at the upper unit.

[0173] According to one embodiment, when the current subblock is in a block matching mode using multiple reference subblocks, each predicted reference offset may be offset information of the subblock most recently created in block matching mode in the current channel or offset information of the subblock created in block matching mode in the channel immediately preceding the current subblock, and each predicted reference offset may be managed differently. For example, a data storage space for storing a reference offset to point to the first subblock and a data storage space for storing a reference offset to point to the second subblock may be defined in different spaces. For example, after the current subblock performs block matching using two reference subblocks, a new reference offset may be stored in both data storage spaces, but after the current subblock performs block matching using one subblock, a new reference offset may be stored only in the first data storage space, and an update with the new reference offset value may not occur in the second data storage space.

[0174] According to one embodiment, when the prediction mode of the current subblock is block matching mode and multiple reference subblocks are used, the prediction reference offset value can be determined in the same way for all reference subblocks. For example, for all prediction reference offsets, whether to use the reference offset of the subblock most recently selected in block matching mode in the current channel or the reference offset of the subblock most recently selected in block matching mode in the previous channel can be determined through a single 1-bit flag signaling / parsing.

[0175] According to one embodiment, when the prediction mode of the current subblock is a block matching mode and multiple reference subblocks are used, the residual offset for each reference subblock can be determined through signaling / parsing for each reference subblock, and in the process, the sign and absolute value of each residual offset can be separated and signaled / parsed.

[0176] According to one embodiment, when signaling / parsing a residual offset for each reference subblock, the i-th residual offset can be determined by signaling / parsing the value obtained by differentiating the first residual offset from the i-th residual offset instead of the i-th residual offset as in Equation 10. Here, u is 2 or greater and may be a value greater than or equal to i.

[0177]

[0178] Here, t diff, i is the residual offset for the i-th subblock, t diff, 1 can represent the residual offset for the first subblock. tdiff_signaled, i can represent the difference value signaled / parsed for the i-th subblock.

[0179] The value of tdiff_signaled, i can be determined through signaling / parsing, and the residual offset for the i-th subblock can be determined by summing the first residual offset for the above parameter. Additionally, when signaling / parsing the value obtained by differentiating the first residual offset from the i-th residual offset, the sign and absolute value can be separated for signaling / parsing.

[0180] According to one embodiment, when the prediction mode of the current subblock is a block matching mode and multiple reference subblocks are used, a 1-bit flag can be signaled / parsed to determine whether the final reference offset for each subblock is determined as the predicted reference offset, or whether the predicted reference offset for each subblock is determined and additionally the residual reference offset for each subblock is signaled / parsed to determine the sum of the predicted reference offset and the residual reference offset.

[0181] According to one embodiment, when the prediction mode of the current subblock is a block matching mode and multiple reference subblocks are used, the final reference offset of each reference subblock can be determined as the predicted reference offset for each reference block, and in the process, a list of reference offsets can be constructed to determine the predicted reference offset.

[0182] According to one embodiment, when the prediction mode of the current subblock is a block matching mode using multiple reference subblocks and a list of reference offsets is constructed, at least one of the previously restored subblock(s) adjacent to the left of the current subblock, the previously restored subblock(s) located at the same position as the current subblock and in the previous channel, and the previously restored subblock(s) located to the left of the current subblock and in the previous channel may be searched to construct the list. During the search process, if a previously restored subblock exists in a block matching mode using multiple reference blocks, all reference offsets of that subblock may be added to the list. Additionally, if the size of the list is full, the search process may be terminated early.

[0183] According to one embodiment, when the prediction mode of the current subblock is a block matching mode using a plurality of reference subblocks and a list of reference offsets is configured, the index of the list for each reference subblock can be signaled / parsed to determine each predicted reference offset, and the predicted reference offset for each reference subblock can be determined as the final reference offset.

[0184] According to one embodiment, in a block matching mode using a plurality of reference subblocks, when a reference offset pointing to each subblock is determined by signaling / parsing an index of a list, an additional correction value for each reference offset can be signaled / parsed to determine the corrected reference offset as the final reference offset. Whether to perform the correction can be determined by signaling / parsing a 1-bit flag.

[0185] According to one embodiment, when signaling / parsing additional correction values ​​for a reference offset, the sign and absolute value of the correction value can be separated for signaling / parsing. The absolute value of the correction value may be predefined in a table form in the encoder / decoder, and the value existing in the table can be modified by signaling / parsing information regarding the absolute value at the waveform parameter level. The absolute value of the correction value can be determined by signaling / parsing the index of the table.

[0186] According to one embodiment, when the prediction mode of the current subblock is a block matching mode and a plurality of reference subblocks are used, a linear model between the template region of the plurality of reference subblocks and the template region of the current subblock can be derived and the linear model applied to the plurality of reference subblocks to generate a prediction signal of the current subblock.

[0187] For example, linear model parameter α from the template regions of u reference subblocks and the current subblock's template region block,1, αblock,2,쪋,α block,u, β blockThe prediction signal for the current subblock can be generated by deriving and applying the corresponding linear model parameters to the reference subblock. For example, the linear model parameter α from the template regions of the two reference subblocks and the template region of the current subblock block,1, α block,2, β block It is possible to generate a prediction signal for the current subblock by deriving the corresponding linear model parameters and applying them to the reference subblock.

[0188] According to one embodiment, linear model parameter α from the template region of a plurality of reference subblocks and the template region of the current subblock block,1, αblock,2,… ,α block,u, β block can be derived, and α block,1, αblock,2,… ,α block,u, β block It can be derived from a method that minimizes the cost (e.g., MSE) between the template region of the current subblock and the template region of multiple referenced subblocks. For example, linear model parameters can be derived through Equation 11.

[0189]

[0190] Here, t1 is the reference offset pointing to the first referenced subblock, t u can represent a reference offset pointing to the u-th referenced subblock. Here, rec[m][-N-t1+q] is the template area of ​​the first referenced subblock, rec[m][-Nt u [+q] represents the template area of ​​the u-th referenced subblock, and rec[m][-N+q] represents the template area of ​​the current subblock.

[0191] According to one embodiment, when deriving linear model parameters from the template regions of a plurality of reference subblocks and the template regions of the current subblock, the linear model derivation process can be simplified by signaling / parsing a 1-bit flag to set the scaling parameter to 1 / u and deriving only the bias parameter. For example, the simplified linear model can be derived through Equation 12.

[0192]

[0193] Here, t1 is the reference offset pointing to the first referenced subblock, t u can represent a reference offset pointing to the u-th reference subblock. Here, the scaling parameter (α block, i ) can be set to 1 / u. Here, u is greater than or equal to 1 and can be a value greater than or equal to i.

[0194] According to one embodiment, a linear model can be derived from the template regions of a plurality of reference subblocks and the template regions of the current subblock, and the derived linear model can be applied to a plurality of reference subblocks to generate a prediction signal of the current subblock. For example, the prediction signal can be calculated through Equation 13.

[0195]

[0196] Here, rec[j-t1], rec[j-t2], rec[jt u ] can represent the first, second, and u-th referenced subblocks, respectively. α block,1 , α block,2 , … , α block,u can represent the first, second, and u-th scaling parameters, respectively. β block can represent bias parameters.

[0197] According to one embodiment, when the prediction mode of the current subblock is a block matching mode and multiple subblocks are used, whether to generate a prediction signal of the current subblock by applying a linear model to multiple reference subblocks or to generate a prediction signal of the current subblock through a weighted sum of multiple reference subblocks can be determined through a 1-bit flag.

[0198] According to one embodiment, when generating a prediction signal of the current subblock as a weighted sum of a plurality of reference subblocks, the prediction signal of the current subblock can be generated as an average of a plurality of reference subblocks. For example, the prediction signal of the current subblock can be generated through Equation 14.

[0199]

[0200] Here, pred 0 [j] and pred 1 [j] can represent a prediction signal generated from different reference subblocks (or different reference offsets), and l k can represent the length of the current subblock.

[0201] According to one embodiment, when generating a prediction signal for a current sub-block by a weighted sum of a plurality of sub-blocks, different weights may be applied to each reference sub-block, and a prediction signal for a final current sub-block may be generated by weighted summing the signals predicted from each reference sub-block. For example, a prediction signal for a current sub-block may be generated through Equation 15.

[0202]

[0203] Here, pred 0 [j] and pred 1 [j] can represent a prediction signal generated from different reference subblocks (or different reference offsets), and l k can represent the length of the current subblock. a and b can represent weights.

[0204] According to one embodiment, when the prediction mode of the current subblock is a block matching mode and a prediction signal of the current subblock is generated using a plurality of subblocks, whether to generate the prediction signal by the average of a plurality of reference subblocks or by generating the prediction signal through a weighted sum can be determined through a 1-bit flag.

[0205] According to one embodiment, when generating a prediction signal of the current subblock as a weighted sum of a plurality of reference subblocks, the weights may be determined by signaling / parsing on a subblock basis, a block basis, or a frame basis. The weights may be determined by signaling / parsing directly or by signaling / parsing an index from a predefined table for weights.

[0206] According to one embodiment, if the prediction mode of the current subblock is a block matching mode using one reference subblock and a linear model is not applied, the average of the difference in sample values ​​between the template area of ​​the reference subblock and the template area of ​​the current subblock can be calculated. Subsequently, the calculated average value can be added to each sample of the reference subblock to generate a final prediction signal for the current subblock. Whether to generate a final prediction signal for the current subblock by adding the average value of the difference in sample values ​​between the two templates to the reference subblock, or to generate a final prediction signal for the current subblock using the reference subblock, can be determined by signaling / parsing a 1-bit flag.

[0207] According to one embodiment, when the prediction mode of the current subblock is a block matching mode using multiple reference subblocks and a linear model is not applied, when generating a prediction signal for the current subblock from each reference subblock, the average of the difference in sample values ​​between the template area of ​​each reference subblock and the template area of ​​the current subblock can be calculated. Subsequently, the calculated average value can be added to each sample of each reference subblock to generate a prediction signal for the current subblock for each reference subblock. Finally, the prediction signals for the current subblock for each reference subblock obtained through the above process can be weighted and summed to generate a final prediction signal for the current subblock. Whether to apply the above process can be determined by signaling / parsing a 1-bit flag.

[0208] According to one embodiment, if the prediction mode of the current subblock is a block matching mode, a final prediction signal can be generated by applying filtering to the prediction signal generated in the block matching mode. At this time, a 1-bit flag can be signaled / parsed to determine whether filtering is performed. If filtering is performed, the filtered signal can be the final prediction signal, and if filtering is not performed, the generated prediction signal can be the final prediction signal.

[0209] According to one embodiment, when performing filtering, a 1-bit flag may be additionally signaled / parsed to determine the type of filter.

[0210] According to one embodiment, filter coefficients for applying filtering can be defined in advance in a table. The table for filter coefficients can be changed on a frame-by-frame or channel group-by-channel basis, and when filter coefficients are changed, each filter coefficient can be determined by directly signaling / parsing.

[0211] According to one embodiment, if the filter coefficients are defined in a table form, the filter coefficients can be determined by signaling / parsing the index of the filter coefficient table.

[0212] Referring to FIG. 8, residual samples for the current subblock can be generated based on residual information for the current subblock (S820).

[0213] According to one embodiment, quantized residual samples in which statistical information is compressed can be parsed. The compressed statistical information can be restored using a variable-length encoding method or an arithmetic encoding method.

[0214] According to one embodiment, residual samples for the current subblock can be restored. During the inverse quantization process, the sign part and the level part of the residual samples can each be restored through a finite state machine based on parity bits.

[0215] According to one embodiment, the conversion and inverse conversion units may be determined as a single sub-block or block.

[0216] According to one embodiment, a transformation kernel for performing an inverse transformation can be determined. The transformation kernel can be determined through signaling / parsing and can be implicitly determined based on subblocks or block sizes, etc.

[0217] According to one embodiment, an inverse transformation can be performed on the transformation coefficients using a transformation kernel.

[0218] Referring to FIG. 8, the current subblock can be restored based on prediction samples and residual samples (S830).

[0219] According to one embodiment, a reconstructed sample can be generated based on a predicted sample and a residual sample. A biomedical waveform signal can be generated from the reconstructed sample.

[0220] FIG. 12 is a diagram showing the schematic configuration of a biomedical waveform signal decoding device according to one embodiment of the present disclosure.

[0221] Referring to FIG. 12, the decoding device (1200) may include a prediction sample generation unit (1210), a residual sample generation unit (1220), and a restoration unit (1230).

[0222] The prediction sample generation unit (1210) can perform a prediction sample generation process according to S810. The prediction sample generation unit (1210) can perform at least one of the following processes: determining a prediction unit, determining a prediction type, determining a prediction mode, and performing a prediction. The order of each process may be changed and may be omitted.

[0223] The residual sample generation unit (1220) can perform a residual sample generation process according to S820. The residual sample generation unit (1220) can perform at least one process among decoding the transformation coefficient entropy, inverse quantization, determining the inverse transformation unit, determining the transformation kernel, and performing the inverse transformation. The order of each process may be changed and may be omitted.

[0224] The restoration unit (1230) can perform a restoration process according to S830.

[0225] FIG. 13 is a flowchart of a biomedical waveform signal encoding method according to one embodiment of the present disclosure.

[0226] Referring to FIG. 13, prediction samples for the current subblock can be generated based on the previously restored reference subblock (S1310).

[0227] The method for generating current subblock prediction samples based on the previously restored reference subblock is as described with reference to S810.

[0228] Referring to FIG. 13, residual samples for the current subblock can be generated based on prediction samples for the current subblock (S1320).

[0229] The method for generating residual samples for the current subblock is as described in reference to S820.

[0230] Referring to FIG. 13, the transformation coefficients of the current subblock can be derived based on residual samples (S1330).

[0231] Referring to FIG. 13, residual information regarding the conversion coefficients can be encoded into a bitstream (S1340).

[0232] The method for deriving the transformation coefficients of the current subblock and encoding the residual information is as described in reference to S830.

[0233] FIG. 14 is a diagram showing the schematic configuration of a biomedical waveform signal encoding device according to one embodiment of the present disclosure.

[0234] Referring to FIG. 14, the encoding device (1400) may include a prediction sample generation unit (1410), a residual sample generation unit (1420), a transformation coefficient derivation unit (1430), and a residual information encoding unit (1440).

[0235] The prediction sample generation unit (1410) can perform a prediction sample generation process according to S1310. The prediction sample generation unit (1410) can perform at least one of the following processes: determining a prediction unit, determining a prediction type, determining a prediction mode, and performing a prediction. The order of each process may be changed or omitted.

[0236] The residual sample generation unit (1420) can perform a residual sample generation process according to S1320.

[0237] The transformation coefficient derivation unit (1430) can perform a transformation coefficient derivation process according to S1330. The transformation coefficient derivation unit (1430) can perform at least one process among determining a transformation unit, determining a transformation kernel, performing a transformation, and quantization. The order of each process may be changed and may be omitted.

[0238] The residual information encoding unit (1440) can perform the entropy encoding process of the residual information according to S1340.

[0239] The various embodiments of the present disclosure are not intended to list all possible combinations but to describe representative aspects of the present disclosure, and the matters described in the various embodiments may be applied independently or in combination of two or more.

[0240] In addition, various embodiments of the present disclosure may be implemented by hardware, firmware, software, or a combination thereof. In the case of implementation by hardware, it may be implemented by one or more ASICs (Application Specific Integrated Circuits), DSPs (Digital Signal Processors), DSPDs (Digital Signal Processing Devices), PLDs (Programmable Logic Devices), FPGAs (Field Programmable Gate Arrays), general processors, controllers, microcontrollers, microprocessors, etc.

[0241] The scope of the present disclosure includes software or machine-executable instructions (e.g., operating system, application, firmware, program, etc.) that enable an operation according to a method of various embodiments to be executed on a device or computer, and a non-transitory computer-readable medium on which such software or instructions, etc. are stored and executable on a device or computer.

[0242] The present invention can be used to encode / decode signals.

Claims

1. A step of generating prediction samples for the current subblock based on the previously restored reference subblock; A step of generating residual samples for the current subblock based on residual information for the current subblock obtained from the bitstream; The method includes the step of restoring the current subblock based on the prediction samples and the residual samples, wherein The above prediction samples are generated based on a first prediction mode based on a reference offset indicating the location of the above reference subblock, a decoding method.

2. In Paragraph 1, The above reference offset is determined as a predicted reference offset, and A decoding method in which a candidate list for the reference offset is constructed when the reference offset is determined to be a predicted reference offset.

3. In Paragraph 2, A decoding method for searching for at least one subblock among a previously restored subblock adjacent to the left of the current subblock, a subblock at the same location as the current subblock and restored in a previous channel, or a subblock at the left of the current subblock and restored in a previous channel to construct the above candidate list.

4. In Paragraph 2, The above candidate list is a decoding method that adds a reference offset of a subblock generated in the first prediction mode as a candidate among subblocks processed before the current subblock according to the coding order.

5. In Paragraph 2, The above reference offset determines a first reference offset based on index information of the candidate list obtained from the bitstream. A decoding method determined by a second reference offset that has been corrected for the first reference offset.

6. In Paragraph 1, The above prediction samples are generated by applying at least one linear model parameter to the above-mentioned restored reference subblock, wherein A decoding method in which at least one linear model parameter is derived to a value that minimizes the cost between the template region of the previously restored reference subblock and the template region of the current subblock.

7. In Paragraph 6, A decoding method in which at least one scaling parameter included in the above at least one linear model parameter is set to 1.

8. In Paragraph 1, The above prediction samples are generated by applying at least one linear model parameter to the above-mentioned restored reference subblock, wherein A decoding method wherein the above-mentioned at least one linear model parameter is derived based on a first average value for p(p≥1) samples having a maximum value among at least one sample in the template region of the reference subblock, a second average value for p(p≥1) samples having a maximum value among at least one sample in the template region of the current subblock, a third average value for p(p≥1) samples having a minimum value among at least one sample in the template region of the reference subblock, and a fourth average value for p(p≥1) samples having a minimum value among at least one sample in the template region of the current subblock.

9. In Paragraph 1, If the previously restored reference subblock includes at least two reference subblocks, the reference offset is determined for each reference subblock, A decoding method in which the reference offset for each of the above reference subblocks is determined by the predicted reference offset.

10. In Paragraph 9, A decoding method in which the above-mentioned prediction reference offset is determined as either the reference offset of the last processed subblock among the subblocks generated in the first prediction mode that precedes the current subblock in coding order within the current channel, or the reference offset of the last in coding order subblock generated in the first prediction mode in the channel immediately preceding the current subblock.

11. In Paragraph 1, The above prediction samples calculate an average value for the difference in sample values ​​between the template area of ​​the previously restored reference subblock and the template area of ​​the current subblock, and A decoding method generated by adding the above average value to each sample of the above restored reference subblock.

12. A step of generating prediction samples for the current subblock based on the previously restored reference subblock; A step of generating residual samples for the current subblock based on prediction samples for the current subblock; A step of deriving transformation coefficients of the current subblock based on the above residual samples; The method includes the step of encoding residual information regarding the above-mentioned transformation coefficients into a bitstream, The above prediction samples are generated based on a first prediction mode based on a reference offset indicating the location of the above reference subblock, in an encoding method.

13. In a digital storage medium for storing a bitstream, The above bitstream generates prediction samples for the current subblock based on the previously restored reference subblock; A step of generating residual samples for the current subblock based on prediction samples for the current subblock; A step of deriving transformation coefficients of the current subblock based on the above residual samples; Encoded by an encoding method comprising the step of encoding residual information regarding the above-mentioned transformation coefficients into a bitstream, A digital storage medium, wherein the above prediction samples are generated based on a first prediction mode based on a reference offset indicating the location of the above reference subblock.

Images