Method and device for encoding and decoding a signal

Abstract

The invention relates to a method for encoding a signal comprising at least two zones and comprising the following steps for one zone (Zoj) comprising a plurality of samples (Pn) to be encoded, constructing (E22) at least one feature map representative of the zone; for at least one sample, referred to as the current sample (Pn), of the zone (Zoj) and associated with a point in the at least one feature map, encoding a data item, referred to as the current value (Vn), associated with the point, comprising evaluating (F221) a proximity criterion of the point with respect to a boundary of the zone, constructing (F221) a vicinity vector (Cn) of the point according to the proximity criterion, and entropy-encoding (F223, F224) the current value according to the vicinity vector (Cn); updating (E23, E26, E27) at least one value of the feature map according to an encoding performance measurement, and encoding the feature map (FMci j) representative of the zone.

Description

[0001] Description

[0002] TITLE: Method and device for encoding and decoding a signal

[0003] Previous technique

[0004] The invention relates to the general field of coding one-dimensional or multidimensional signals. It relates more particularly to the compression of digital images or videos.

[0005] Digital videos are generally encoded using source code to compress them and reduce the resources required for transmission and / or storage. Numerous encoding standards exist, such as those of the ITU / MPEG organizations (H.264 / AVC, H.265 / HEVC, H.266 / VVC, etc.) and their extensions (MVC, SVC, 3D-HEVC, etc.). In these approaches, image encoding is typically performed by predicting pixels using previously encoded and decoded pixels present in the image being encoded—in which case it is called "intra-prediction"—or previously encoded images—in which case it is called "inter-prediction."

[0006] In addition to these traditional approaches, approaches based on artificial intelligence, and in particular neural networks, are tending to develop.

[0007] Some neural network approaches use an input signal, such as an image, to train a synthetic neural network on characteristic vectors associated with a position of a sample of the input signal to be encoded. These characteristic vectors are constructed from feature maps that may have the same resolution as the input signal or a lower resolution. During training, or construction, the neural network parameters and the feature map values ​​are updated based on a performance metric, such as bitrate-distortion. When training is complete, meaning the performance metric is satisfactory, the actual encoding of the synthetic neural network parameters and feature map values ​​can be performed and stored or transmitted to the decoder.The decoding of the coded signal is then carried out by applying the synthesis neural network to the feature maps.

[0008] One drawback of classical and previously described neural approaches is that they do not allow the creation of independently decodable areas in the coded signal, which limits not only the ability to interact with the semantic content of the coded signal but also the possibility of limiting the memory required at the decoder level or parallelizing the decoding of this coded signal.

[0009] Therefore, there is a need for a solution that improves upon the classical and neural approaches described above.

[0010] Summary of the invention

[0011] The invention relates to a method for encoding a signal comprising at least two zones, comprising the following steps for a zone comprising a plurality of samples to be encoded:

[0012] - the construction of at least one representative characteristics map of said area,

[0013] - for at least one sample, called the current sample, of said area, associated with a point in said at least one feature map, the coding of a data associated with said point, called the current value, comprising:

[0014] - the evaluation of a proximity criterion for said point in relation to a boundary associated with said zone,

[0015] - the construction of a neighborhood vector of said point as a function of said proximity criterion, and

[0016] - an entropic encoding of said current value from said neighborhood vector,

[0017] - the updating of at least one value of said characteristic map, based on a coding performance measurement,

[0018] - the coding of said characteristic map representing said area.

[0019] The invention also relates to a method for decoding a data stream representative of a signal segmented into zones, said decoding method comprising the following steps, for one zone:

[0020] - the decoding of at least one characteristic map representative of said area, the decoding including, for at least one point on said map:

[0021] - the evaluation of a proximity criterion for said point in relation to a boundary associated with said zone,

[0022] - the construction of a vector, called the neighborhood vector of said point, as a function of said proximity criterion, and

[0023] - an entropic decoding of a data point associated with the point, called the current value, as a function of said neighborhood vector,

[0024] - the application of a neural network to at least one decoded feature map. For the purposes of this invention, encoding, or "coding," means the operation of representing one or more values ​​or a set of samples in a compact form, for example, using a digital binary stream. Decoding means the operation of processing a digital binary stream to recover one or more decoded values ​​or samples.

[0025] A "sample" of the signal refers to a value taken from the signal. Sampling the signal produces a sequence of discrete values ​​called samples. In the case of an image signal, the sample is called a pixel, which can be, for example, a color pixel traditionally represented by a triplet of values, such as (R, G, B) or (Y, U, V). The position of the sample is identified by its x-coordinate (abscissa) and y-coordinate (ordinate) in the image.

[0026] A "signal comprising a plurality of samples" is defined as a signal with one (audio, sound), two (image), or more than two (stereoscopic, multiscopic, depth-mapped, video, etc.) dimensions. Depending on its dimensionality, the sample has one, two, or more coordinates within the signal. In the case of an image signal, the sample's position is identified by its x-coordinate (abscissa) and y-coordinate (ordinate).

[0027] By "signal area," we mean a set of signal samples. For example, in the case of an audio signal, an area might correspond to a succession of units representing silences, noises, phonemes, words, etc. Similarly, in the case of an image signal, an area might correspond to a set of pixels, grouped according to criteria such as intensity, spatial distribution, etc. For example, a segmentation operation can identify two different areas within the image signal, one corresponding to the background and the other to the foreground.

[0028] By "feature maps" or equivalently by "latent value maps", we mean an abstract representation of the signal comprising a plurality of variable data, discrete or not, which are also called values, for example real or integer numbers.

[0029] By "characteristic data vector constructed from feature maps based on a position" we mean a vector consisting of one or more elements, or data, preferably discrete, the data being constructed from the feature maps at a position determined by that of the sample being processed in the signal. This characteristic vector is the one that is applied to the input of the synthesis neural network. For example, in the case of a one-dimensional audio signal, such a vector can be constructed from a plurality of values ​​taken from each of the feature maps at the same coordinate as the sample to be encoded or in a neighborhood of it. In the case of an image, such a vector can be constructed from a plurality of values ​​taken from each of the feature maps at the same x- and y-coordinates as the sample to be encoded (respectively, x, y, and z).to decode) or in a neighborhood of it. Once these values ​​are taken from the feature maps, they can be processed to constitute the feature vector, before entering the synthesis neural network, for example by quantization, filtering, interpolation, etc.

[0030] By "synthetic neural network", we mean a neural network such as a convolutional neural network, a multilayer perceptron, an LSTM (for "Long Short Term Memory"), etc. The neural network is defined for example by a plurality of layers of artificial neurons and by a set of activation, weighting and addition functions (for example, a layer can calculate y = f (Ax+b), where y and b are vectors of dimension N, x a vector of dimension M, A is a matrix of dimension MxN, and f is the activation function).

[0031] By "neural network parameter" we mean one of the values ​​that characterizes the neural network, for example a weight associated with one of the neurons (filter coefficient, weighting, bias, value affecting the functioning of non-linearity, etc.)

[0032] "Processing by a synthetic neural network" refers to the application of a function expressed by a synthetic neural network to the input characteristic vector to produce an output vector representative of the sample to be encoded (or decoded). This output vector may contain one or more data points representative of the sample.

[0033] A "performance measure" is defined as a measurement between at least one value of a sample to be encoded and a decoded value of that sample. The measure might assess, for example, distortion or perceptual error. It can be performed on one or more samples (e.g., a running sample, the current image, etc.). The measure may also include a measurement of throughput, particularly associated with the encoding of the synthetic neural network and / or the encoding of the feature maps in the first group. The measure may be a combined measurement of throughput and distortion through their weighting. As is well known in the state of the art, the value of this measure is generally minimized until a target value is reached.

[0034] Generally, the steps of an encoding or decoding process should not be interpreted as being linked to a notion of temporal succession. In other words, the steps can be carried out in a different order than that indicated in the independent encoding or decoding claim, or even in parallel.

[0035] The coding method according to the invention encodes a region of a signal from a representation of that signal in the form of feature maps. These feature maps are segmented into characteristic regions corresponding to areas of the signal, the values ​​of which are subsequently entropically encoded independently of one another. Thus, it is possible to obtain a coded representation of a region of the original signal that can subsequently be decoded independently of any other part of that signal.

[0036] The decoding process (and symmetrically the encoding process) may further include one or more of the following optional features, taken individually or in any technically possible combination.

[0037] According to a first characteristic, the coding (resp. decoding) also includes a coding (resp. decoding) step of a set of parameters representative of said synthetic neural network.

[0038] According to another characteristic, the encoding (resp. decoding) includes an encoding (resp. decoding) step of segmentation information to determine, in said at least one feature map, a feature area corresponding to said area.

[0039] Segmentation information refers to any information that identifies the different areas requiring specific synthesis during decoding (or encoding). For example, in the case of an image signal, segmentation information indicates a division of the image into a plurality of regular areas of identical size and shape: the image can be divided into rectangular areas with horizontal and vertical boundaries, or into areas corresponding to samples (or blocks) processed in a lexicographical order. In this case, the segmentation information can indicate the beginning and / or end of the set. As another example, a region can be of arbitrary size and shape. In this case, the segmentation information can indicate the shape and size of the region, in the form of a segmentation mask.

[0040] According to another characteristic, the neighborhood of said point is independent of said evaluation of the proximity criterion.

[0041] According to another characteristic, the neighborhood of said point is selected based on said evaluation of the proximity criterion.

[0042] According to another characteristic, during the construction of the neighborhood vector, a component of the neighborhood vector is associated with a point of said neighborhood, the value of said component being equal to the value of said associated point if said associated point is a point of said characteristic zone, and to a replacement value otherwise.

[0043] According to another characteristic, the replacement value is dependent on the points of said characteristic zone or the points associated with said neighborhood belonging to said characteristic zone.

[0044] According to another characteristic, the replacement value does not depend on the points of said characteristic area.

[0045] According to another characteristic, the neighborhood of said point is selected based on said point.

[0046] According to another characteristic, the construction of the neighborhood vector includes: the identification of the points of said neighborhood belonging to said characteristic zone, called identified points, the initialization of the number of components of said neighborhood vector to the number of identified points, and the assignment of the value of each identified point to a different component of said neighborhood vector.

[0047] According to another characteristic, the criterion of proximity of said point in relation to a boundary of an area is a distance.

[0048] According to another characteristic, the neighborhood vector is used to determine a context for the entropic decoding of said current value.

[0049] According to another feature, the entropic decoding of said current value includes a step of predicting said current value as a function of said neighborhood vector.

[0050] According to another characteristic, the decoding of said current value of one of said latent value maps includes: the decoding of a set of parameters representative of a neural network, called a prediction neural network, and the processing of said neighborhood vector by the prediction neural network, to provide a prediction of said current value.

[0051] Correspondingly, the invention also relates to a signal coding device comprising at least two zones, said coding device comprising a processor coupled to a memory in which are stored instructions configured to implement, for a zone of said signal comprising a plurality of samples to be coded: the construction of at least one characteristic map representative of said zone, for at least one sample, called the current sample, of said zone, associated with a point in said at least one characteristic map, the coding of a data associated with said point, called the current value, comprising: o the evaluation of a proximity criterion of said point with respect to a boundary of said zone, o the construction of a neighborhood vector of said point as a function of said proximity criterion, and o an entropic coding of said current value as a function of said neighborhood vector,the updating of at least one value of said feature map, based on a coding performance measure, the coding of said feature map representative of said area.

[0052] The invention also relates to a device for decoding a data stream representative of a signal segmented into zones, the device comprising a processor coupled to a memory in which are stored instructions configured to implement the following steps, for a zone: the decoding of at least one feature map representative of said zone, the decoding comprising, for at least one point of said map: o the evaluation of a proximity criterion of said point with respect to a boundary of said zone, o the construction of a neighborhood vector of said point as a function of said proximity criterion, and o an entropic decoding of a data associated with the point, called current value, from the neighborhood vector, the application of a neural network to said at least one decoded feature map.

[0053] The features and advantages of the encoding or decoding method apply equally to the encoding or decoding device according to the invention, and vice versa. The invention also relates to a computer program on a recording medium, this program being capable of being implemented in a computer or an encoding or decoding device according to the invention. This program includes instructions adapted for implementing the corresponding method. This program may use any programming language and may be in the form of source code, object code, or code intermediate between source and object code, such as in a partially compiled form, or in any other desirable form.

[0054] The invention also relates to a computer-readable information or recording medium containing the aforementioned computer program instructions. The information or recording medium can be any entity or device capable of storing programs. For example, the medium may include a storage means, such as a ROM (e.g., a CD-ROM or a microelectronic circuit ROM), or a magnetic recording means, such as a floppy disk or hard drive, a DNA sequence, or flash memory. Furthermore, the information or recording medium may be a transmissible medium, such as an electrical or optical signal, which can be transmitted via an electrical or optical cable, by radio link, by wireless optical link, or by other means.

[0055] The program according to the invention can in particular be downloaded onto an Internet-type network.

[0056] Alternatively, each information or recording medium can be an integrated circuit in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of a process according to the invention.

[0057] Brief description of the figures

[0058] The invention will be better understood with the aid of the following description, given solely by way of example and made with reference to the accompanying drawings in which: Figure 1 schematically represents a coding device according to a first embodiment of the invention; Figure 2 schematically represents a decoding device according to the first embodiment of the invention; Figure 3 illustrates an example of a synthetic artificial neural network used in the context of the invention; Figure 4 illustrates an example of a predictive artificial neural network used in the context of the invention; Figure 5 is a flowchart representing an example of a coding method that can be implemented by the coding device of Figure 1; Figure 6 illustrates a coding method used in an embodiment of the invention.Figure 7 is a flowchart representing an example of a decoding process that can be implemented by the decoding device of Figure 2; Figure 8 illustrates a decoding process used in an embodiment of the invention; Figure 9 is a flowchart representing a method for encoding feature cards that can be implemented by the encoding device of Figure 1 and by the encoding process of Figure 5.Figure 10 is a logic diagram representing a method for decoding feature maps that can be implemented by the decoding device in Figure 2 and by the decoding process in Figure 7. Figures 11 and 12 illustrate the formation of a neighborhood vector used in a method for encoding or decoding feature maps. Figure 13 illustrates a first variant for forming the neighborhood vector. Figure 14 illustrates a second variant for forming the neighborhood vector. Figure 15 illustrates a third variant for forming the neighborhood vector. Figure 16 illustrates a fourth variant for forming the neighborhood vector.

[0059] Detailed description of the invention

[0060] Figure 1 schematically represents, according to a first embodiment, an ENC coding device for at least one region of a signal (I(P n)). In the example described here, the signal I(P n The system is segmented into a set of J zones, and all zones are coded independently. Alternatively, only one zone or only a few zones can be coded independently.

[0061] This ENC coding device includes a SEG segmentation module, an SEGC encoding module of the segmentation provided by the SEG segmentation module, a GEN module for generating feature maps, an SE transformation module, an XTR data extraction module, a TT processing and quantification module, an NNSYN module corresponding to a synthetic artificial neural network, an NNC neural network coding module capable of encoding the synthetic neural network and optionally the prediction neural network, an FMC feature map coding module, an EVAL module for evaluating coding performance, and an MAJ update module.

[0062] The ENC coding device can be implemented by means of an electronic device comprising a processor and a memory, not shown; each of the modules mentioned above can then be realized by the cooperation of the processor and computer program instructions stored in the aforementioned memory and designed to perform the functionalities of the module concerned, in particular as described below, when these instructions are executed by the processor.

[0063] The ENC coding device in Figure 1 receives as input a signal consisting of a succession of samples to be coded, denoted P n , for example a temporal sequence of sound samples, or a set of image data denoted I(P n ). In this second case, the image signal I(P n) can represent a two-dimensional image, or a plurality of two-dimensional images (video, color components, stereoscopic components, multiscopic components, etc.). P n denotes a sample n of the input signal comprising N samples. In one embodiment, the signal is a color image signal represented by means of at least one two-dimensional representation, such as a pixel matrix, each pixel comprising a red (R), green (G), blue (B) component, or, alternatively, a luminance (Y) component and at least one chrominance (U, V) component.

[0064] The location of each pixel is defined by its x and y coordinates in the image. In one embodiment, the image is a greyscale representation using a two-dimensional representation, such as a pixel matrix, with each pixel having a greyscale component, or luminance.

[0065] The SEG module performs a segmentation Sg of the sequence of samples to be coded Pn into J (greater than or equal to 2) zones Zo 1This segmentation operation allows the samples to be grouped into J different homogeneous sets according to one or more predefined criteria. For example, in the case of a temporal sequence of sound samples, the segmentation operation allows us to obtain sequences of units corresponding to silences, noises, phonemes, words, etc. Similarly, in the case of an image signal I(Pn), the segmentation operation allows the pixels Pn of the image signal to be grouped into J homogeneous zones according to criteria, notably intensity or spatial ones. For example, the segmentation operation can allow us to identify two different zones (J=2) in the image signal I(Pn), one corresponding to the background of the image and the other to the foreground. Other examples of segmentation will be presented later with reference to the figure. The segmentation Sg, for example, is represented as a set of J masks MZo 1allowing each to identify a Zo zone, each of the masks being associated with a different value than that associated with the other masks.

[0066] Optionally, the SEGC module performs lossless encoding of the Sg segmentation information. This encoding can be achieved by encoding the segmentation map corresponding to the J MZo masks. J (for example, by using the JPEG-LS algorithm defined by the international standard ISO / IEC 14495-1) or, alternatively, by independently encoding each MZo' mask, or, in yet another alternative, by independently encoding the contours of the J MZo' masks J , or in yet another variant by encoding a size or position data for an area, for example rectangular. The coded segmentation is denoted Sgc.

[0067] The NNC module performs a simulation of coding, followed by decoding, of the parameters of the NNSYN neural network for the evaluation module.

[0068] The GEN feature map generation module is configured to generate a plurality of M feature maps, denoted LM!.

[0069] In one embodiment, the SE module performs a transformation of the first group of FM! characteristic maps to generate a second group of FMSi characteristic maps at the same resolution as the input signal.

[0070] The optional SE module can perform quantization of the data extracted from this set of M FM! cards. Recall that quantizing a value refers to mapping that value to a member of a discrete set of possible code symbols. For example, the set of possible code symbols might consist of integers, and the quantization system would simply round a real value to an integer. Alternatively, quantization could involve multiplying by a given value and then rounding. The SE module then performs a transformation of the values ​​in at least one of the feature cards, such as oversampling, interpolation, filtering, etc. After the transformation, a transformed feature card from the second group has the same resolution as the images in the input sequence.Advantageously, according to this method, the feature maps that are coded can be of lower resolution than the images to be coded, while the maps of the second group, which are used to construct the feature vectors, are at the same resolution as the image sequence, which facilitates the extraction of values.

[0071] In one embodiment, the SE module is absent; in this case, the values ​​that will be used to construct the characteristic vector are extracted from the first group of characteristic maps.

[0072] The XTR module performs value extraction from the FMSi (or FM!, depending on one of the previously described embodiments) feature maps, for a current sample P n to be encoded, depending on its coordinates in the input signal. For example, if we want to encode sample P n at coordinates (x n , y nFrom an input image, the XTR module performs an extraction of values ​​in the maps at positions imposed by the coordinates (x n , y n ).

[0073] In one embodiment, the extracted values ​​constitute the vector Z n . Z n is an L-tuple, meaning it contains L elements, or data Zi. For example, in one embodiment, L=M, meaning that only one value is extracted for each feature map FM!. In another embodiment, L>M, meaning that several values ​​are extracted for at least one feature map FMi. The dimension L of the vector depends on the topology of the NNSYN synthesis neural network and, more specifically, on the topology of the input layer of this NNSYN synthesis neural network.

[0074] The vector Z n index n refers to the characteristic vector of pixel P' n .

[0075] In one embodiment, the optional TT module processes the extracted values ​​to generate the vector Z n The TT module can perform quantification of the data extracted from the feature set. The processing may include other operations, such as filtering, scaling, etc. In particular, if the SE module is not used and if the feature maps in the first group have lower resolutions than the images in the sequence, the TT module can take into account the coordinates of the values ​​in the lower-resolution maps.

[0076] Note that at least one of the SE or TT modules must perform a quantification of the feature maps.

[0077] The NNSYN module is a synthetic neural network defined by K parameters Wk, capable of processing the vector Z n, or L-tuple, as input, to generate as output a second vector representative of the sample P n to be coded. The synthetic neural network, in one embodiment, is formed from an MLP, or Multi-Layer Perceptron, consisting of an input layer adapted to the input format (the L-tuple), optionally one or more hidden layers, and an output layer adapted to the output format of the output vector, generally a vector containing A elements. In one embodiment, A is equal to 3 and the output vector is the triplet (R, G, B) of the pixel P'. n encoded then decoded.

[0078] The NNC module performs the coding of the synthesis neural network, specifically its parameters Wk. Optionally, the NNC module also performs the coding of the prediction neural network, ARM, specifically its parameters Ob. During the coding training or construction process—that is, until the performance evaluation stage is satisfactory—the NNC module performs a coding simulation, followed by decoding, for the evaluation module. Subsequently, it performs the actual coding of the synthesis neural network parameters Wk, and optionally, the coding of the prediction neural network, ARM. The coded parameters are denoted Wck and Ocb. As is known, the coding simulation can be identical to the actual coding, or it can approximate it.

[0079] The FMC module performs the encoding of the FM maps, that is, the values ​​of the characteristic maps in the first group (excluding the maps in the second group, which may optionally result from oversampling by the SE module). During the encoding training, or construction, process—that is, until the performance evaluation step is satisfactory—the FMC module performs a coding simulation, followed by decoding, for the evaluation module. Subsequently, it performs the actual encoding of the FM map values.

[0080] The FMC module determines, by taking into account the Sg segmentation performed by the SEG module, an SgL segmentation of each of the FM! feature maps into J ZoL zones. Thus, the FMC module uses the Sg segmentation defined in the domain of the signal to be encoded to obtain a segmentation in the latent domain of the FM! feature maps. Obtaining this segmentation in the latent domain depends on the transformation transform used to go from the domain of the signal to be encoded to the latent domain. Therefore, the SgL segmentation in the latent domain results from a calculation that simultaneously considers the location of the Sg segmentation zones in the signal to be encoded and the transformation transform. For example, if the latent domain and the signal to be encoded have the same resolution, then the segmentation in the latent domain is identical to the segmentation in the domain of the signal to be encoded.According to another example, if a latent has a resolution lower than that of the signal to be encoded, then the boundary between two zones of the segmentation in the domain of this latent lies between the points in the latent domain whose colocalized samples, i.e., corresponding samples in the signal to be encoded, belong to two different zones of the Sg segmentation. Each of the FM-"' maps is represented as a set of J zones ZoLj, each of the ZoL zones. J j corresponds in the FM! map to region j of the SgL segmentation. The encoding of the FM! map values ​​is performed taking into account the SgL segmentation. More precisely, each of the J ZoL zones J r iis coded sequentially, for example, in the order of their indexing. The coding of a ZoL zone involves the successive coding, for each FMi map, of the point values ​​of the ZoL'j zone in the FMi map. Thus, each ZoL zone is coded independently of the other zones. The coded data of each coded zone Zol,', in each of the FMi feature maps are denoted FMcJ. Each Zol,' zone undergoes entropic coding. Thus, the coding of the M FMi maps corresponds to the independent coding of the J zones of the M FMi maps.

[0081] As is known, the coding simulation can be identical to the actual coding, or it can approximate it. The coding module quantizes, if necessary, the latent representation of the values ​​in the first group of maps using a quantizer to generate an ordered collection of quantized values. The coding module then compresses the quantized data, using a coding method that takes into account the neighborhood of a value to be coded in the feature map. As will be detailed later, the FMC module can include a predictive neural network (PNN).

[0082] The EVAL module performs an evaluation and minimization of coding performance. The evaluation function is, for example, of the rate-distortion type. Minimization can be performed by gradient descent, or any other method within the grasp of a person skilled in the art.

[0083] The MAJ module performs an update of the values ​​of the FMi cards and / or the parameters of the neural network(s) to be encoded, according to the results of the performance function.

[0084] Figure 2 schematically represents a DEC decoding device for a zone to be decoded Zod J of a signal, called the decoding zone, said decoding zone Zod J comprising a plurality of Pd samples n to decode.

[0085] This decoding device (DEC) comprises an NND module for decoding neural networks capable of decoding the NNSYN' synthesis neural network and optionally the ARM' prediction neural network, a SEGD segmentation decoding module, an FMD feature map decoding module, an XTR' data extraction module, an SE' inverse transformation module, and a TT' inverse processing and quantization module. The DEC outputs a decoded image containing at least the decoded region, denoted Zod. J (Pd n ), comprising a plurality of decoded samples Pd n .

[0086] The DEC decoding device in Figure 2 receives as input the Sgc encoded segmentation and a group of FMcJ encoded data.

[0087] The DEC decoding device in Figure 2 also receives as input the Wck encoded parameters of the NNSYN' synthesis neural network and optionally the Ocb encoded parameters of the ARM' synthesis neural network.

[0088] The parameters of the NNSYN' synthesis neural network decoded by the NND module are denoted Wdk. The parameters of the ARM' prediction neural network decoded by the NND module are denoted Odb.

[0089] The SEGD module optionally decodes segmentation information, for example the MZo\ mask from Sgc encoded data.

[0090] According to embodiments as described for the encoder:

[0091] The segmentation map corresponding to the J masks MZo 1 is decoded and the MZo mask J is extracted from the segmentation map.

[0092] The MZo mask Jis decoded directly when it has been previously encoded in Sgc-encoded data independently of other masks.

[0093] If the segmentation information is not decoded, it can be inferred by the decoder (for example, because it is static, or because the decoder knows the size of the rectangular areas corresponding to the tiles or the number of blocks in the slices, etc.)

[0094] The FMD module constructs M decoded FMdJ maps using the MZo mask 1 and the FMcJ-encoded data. From the MZo mask 1 In the domain of the signal to be decoded, the FMD module determines the ZoL zone J corresponding to the latent domain, then the FMD module decodes the FMcJ data to obtain the point values ​​of this ZoL zone. J in each of the FMdJ maps.

[0095] In one embodiment, the SE' module transforms the first set of decoded FMdJ feature maps to generate a second set of feature maps with the same resolution as the signal to be decoded, denoted FMS'J. Optionally, the SE' module performs inverse quantization corresponding to the quantization performed by the encoder. Inverse quantization is not necessary if the encoder's quantizer Q simply rounded the input values. Inverse quantization is also unnecessary if the neural network is capable of handling quantization of its input data. Otherwise, the decoder performs the inverse operation of quantizer Q. The SE' module then transforms the feature map values, including, for example, oversampling, interpolation, filtering, etc., similar to the transformation performed by the encoder.At the end of the transformation, a transformed feature map of the second group has the same resolution as the images of the sequence to be decoded.

[0096] In one embodiment, the SE' module is absent; in this case, the values ​​that will be used to construct the characteristic vector are extracted from the first group of characteristic maps.

[0097] The XTR' module is identical to the XTR module in Figure 1. It performs an extraction of values ​​from the M FMdJ (or FMS'J, depending on one of the embodiments described previously) feature maps, for a sample Pd n to be decoded, according to its coordinates in the signal to be decoded.

[0098] In one embodiment, the extracted values ​​constitute the vector Zd n . Zd n is an L-tuple, that is to say, it contains L elements, or zd data,

[0099] In one embodiment, the optional TT' module processes the extracted values ​​to generate the vector Zd n The TT module can perform inverse quantization of the data extracted from the feature set. The processing can include other operations, such as filtering, scaling, etc., similar to those performed by the encoder.

[0100] The NNSYN' module is a so-called synthetic neural network, defined by K parameters Wdk, capable of processing the vector Zd n , or L-Tuple, as input, to generate as output a second vector representative of the sample Pd nThe decoding process generally involves a vector containing A elements. In one embodiment, K=3 and the output vector is the triplet (R, G, B) of the decoded pixel Pdn. The NNSYN' module has the same structure as the NNSYN module, and its parameters are either identical if the encoding of its parameters Wk is lossy, or different if the encoding is lossy.

[0101] When all the samples Pd n The signal has been decoded, we have a reconstructed Zod signal J (Pd n ) of the Zo zone J .

[0102] Furthermore, the DEC decoding system can be successively implemented to decode all Zod zones J in order to reconstruct all the samples to be decoded from the signal, i.e., from the image I(Pd nThe DEC decoding device can be implemented using an electronic device comprising a processor and memory (not shown). Each of the modules mentioned above can then be implemented through the cooperation of the processor and computer program instructions stored in the aforementioned memory and designed to perform the functionalities of the module in question, particularly as described below, when these instructions are executed by the processor. The DEC device can also include multiple processors, with each processor dedicated to the parallel decoding of image areas.

[0103] Figure 3 illustrates an example of a synthetic artificial neural network used for encoding and decoding in one embodiment of the invention.

[0104] The synthetic artificial neural network used for encoding, NNSYN, and the synthetic artificial neural network used for decoding, NNSYN', are defined by an identical structure, including for example a plurality of layers of artificial neurons, and by a set of weights and activation functions associated respectively with the artificial neurons of the network concerned.

[0105] A vector representation of a current sample (a vector Z) n or Zd n (derived from the FMi / FMSi feature maps at the encoder or FMdJ / FMS'J at the decoder) is applied to the input (i.e., to an input layer) of the NNSYN or NNSYN' synthetic artificial neural network. The synthetic artificial neural network produces as output a vector, according to one embodiment, the color components (R, G, B) constituting a color pixel of an image or an area of ​​an image.

[0106] The concatenation of all these reconstructed pixels in an image constitutes the image or the decoded, or reconstructed, area of ​​the image.

[0107] At the encoder, the NNSYN synthetic artificial neural network is trained on the entire image, so as to minimize the differences between the input representation of the current image I(P n ) and its representation I(P' n ) at the output, while also minimizing the amount of data to be encoded. The EVAL module performs a performance measurement in this regard.

[0108] Once the encoder training is complete, the network parameters are encoded, either losslessly, in which case the NNSYN' neural network is identical to NNSYN, or lossily, in which case the NNSYN' network may be slightly different from NNSYN.

[0109] Figure 4 illustrates an example of an artificial neural network used for prediction in encoding (ARM) and decoding (ARM') of feature maps, within the framework of embodiments of the invention. The artificial neural network used for prediction in encoding, ARM, and the artificial neural network used for prediction in decoding, ARM', are defined by an identical structure, comprising, for example, a plurality of layers of artificial neurons, and by a set of weights and activation functions associated respectively with the artificial neurons of the network concerned.

[0110] A vector representation of a current neighborhood (a vector C n or CD n derived from the FM! characteristic maps at the encoder or FMdJ at the decoder) is given as input (i.e. on an input layer) of the ARM (at the encoding) or ARM' (at the decoding) artificial neural network for prediction.

[0111] The artificial predictive neural network behaves like a function that outputs a prediction of the current value of the feature map being processed, which can be in the form of a predicted value or probability data.

[0112] According to one embodiment, at the encoder, the network implements a function f v which provides an expected mean and / or variance (p,o) for the current value of the current value V nof the FM card! to be encoded. These statistical values ​​are used to perform entropic coding of this value. For example, if the function produces a mean, this mean is subtracted from the current value, and only the difference is entropically coded, this mean being considered a prediction of the current value. Alternatively, if the function produces a mean and a variance, the mean is subtracted from the current value, and the difference is coded using an entropic coding adapted to the variance, for example, by quantizing said variance into a predetermined set of variances and associating a type of entropic coding with each quantized variance value. At the decoder, the network implements a function f v which provides an expected mean and / or variance for the current value of the current value Vd nof the FMdJ map to be decoded. These statistical values ​​are used to perform entropic decoding of that value. For example, if the function produces a mean, the current value is decoded by the entropic decoder, and this mean is added to the current value. Alternatively, if the function produces a mean and a variance, the current value is decoded by the decoder using an entropic decoding method adapted to the variance, for example, by quantizing the variance into a predetermined set of variances and associating a type of entropic decoding with each quantized variance value.

[0113] In another embodiment, the neural network can produce the expected probability (pr) of each possible value in the current sample. In this case, entropic encoding or decoding will be tailored to this probability (as is known for Huffman or arithmetic entropic coding).

[0114] At the encoder, the ARM artificial neural network for prediction is trained on the entire image, so as to minimize the size (in bytes or bits) of the encoded data. The EVAL module performs a performance measurement in this regard. Recall that the overall performance measurement focuses on minimizing the distortion between encoded and decoded images I(P' n ) and the input images I(P n ), while minimizing the encoding rate. In one embodiment, the feature maps are losslessly encoded using entropy coding. In this case, the encoding of the feature maps affects the data rate but not the distortion of the encoded images. In another embodiment, if the feature maps are lossily encoded, the encoding of the feature maps affects both the data rate and the distortion.

[0115] Once training is complete, the B parameters Ocb of the network are encoded, either losslessly, in which case the neural network ARM' is identical to ARM, or lossily, in which case the network ARM' may be slightly different from ARM.

[0116] Figure 5 is a flowchart representing an example of a coding process that can be implemented by the coding device of Figure 1 when the NNSYN synthesis neural network is the one shown in Figure 3.

[0117] According to this embodiment, the signal is a two-dimensional image; each sample to be coded is therefore a pixel P n coordinates (x n , y n ).

[0118] The encoding process takes place in three main phases:

[0119] In a first phase, called the segmentation phase, the segmentation Sg of the input signal I(P n ) in J (greater than or equal to 2) Zo zones J is done.

[0120] In a first example of segmentation, the image is divided into J regular zones, for example, of identical size and shape (except possibly at the edges of the image). For example, the image can be divided into rectangular zones with horizontal and vertical boundaries. Such a division corresponds to the concept of "tiles" implemented by encoding standards such as H.264 / AVC, H.265 / HEVC, and H.266 / VVC. The segmentation information allows these rectangular zones to be identified, for example, by specifying their vertical and horizontal dimensions.

[0121] In a second example of segmentation, the image is divided into J zones corresponding to samples (or blocks) scanned in a lexicographical order. This division corresponds to the concept of "slices" implemented by encoding standards such as H.264 / AVC, H.265 / HEVC, and H.266 / VVC. Segmentation information allows these zones to be identified, for example, by indicating the number of blocks.

[0122] In a third segmentation example, the image is divided into semantic zones, such as a background and a foreground. This can be done manually by an operator. Segmentation can also be automatic or semi-automatic, depending on the segmentation algorithm used. It's important to note that there are no restrictions related to the type of segmentation algorithm used. The segmentation information allows these zones to be identified, for example, through a segmentation mask.

[0123] Advantageously, such a division into zones allows, during the subsequent decoding of the I(Pdn) signal, for the decoding of the zones to be decoded Zod to be parallelized. J by distributing the decoding load evenly across each of the decoders and / or processors.

[0124] In a second phase, called the construction phase, a learning process is carried out in order to determine, for an input signal I(P nThe values ​​of the LM! maps and the Wk and optionally Ob parameters are used to optimize an overall cost function. Learning is performed, for example, by gradient descent, followed by updating the parameters of the NNSYN synthesis neural network, the FM feature map values, and optionally the ARM prediction neural network. As is known in the state of the art, the cost function can be rate-distortion, rate, distortion, or perceptual. To measure the rate R, it is necessary to simulate the encoding of the J-zones ZoL. JFM-" cards are used, and the associated coding rate (the size of the B2 stream) is then measured. In one embodiment, the coding of parameters Wk and / or Ob is not simulated because their influence is less significant than that of the characteristic maps. In another embodiment, the coding of parameters Wk and / or Ob is also simulated, and the associated rate (the size of the B1 stream) is measured. To measure the distortion D, it is necessary to simulate the coding and then the decoding of at least a portion of the image I, to obtain at least one pixel P'. n resulting from a simulation of encoding and then decoding, then measuring the difference between this part of the image I(P n ) as input and a corresponding part of the image I(P' n ) encoded then decoded.

[0125] Then, during a third phase, called the coding phase, the data from each ZoL zone J r iOptionally, the segmentation Sg, parameters Wk, and possibly Ob are encoded to produce the encoded values ​​Sgc, FMc, and Wck (and possibly Ocb) before transmission or storage. These constitute the compressed representation of the input signal I(P n This compressed signal can be decoded zone by zone. We will now describe the steps of a coding process according to one embodiment of the invention.

[0126] During an E20 step, a signal I(P n ) to be coded, comprising a plurality of N samples P n is provided as input to the process.

[0127] During step E21, the segmentation Sg of the input signal I(P n ) in J (greater than or equal to 2) Zo zones 1 is done.

[0128] During step E22, the M FM! maps of the first group. Subsequently, the Wk parameters of the NNSYN synthesis neural network and the values ​​of the FM! maps, and possibly the Ob parameters of the prediction neural network, must be optimized during the construction phase.

[0129] According to one embodiment, the FMi cards have the same resolution as the input signal I(P n ) and therefore each have the same number of values ​​N as there are samples P n to code.

[0130] According to one embodiment, the FMi cards have a resolution less than or equal to that of the input signal I(P n ) and therefore include, for at least one of them, a number N' of values ​​to be coded less than N. According to a variant, the first FMi card is at the resolution of the images and each subsequent card is at half the resolution of the previous one.

[0131] According to one embodiment, several FMi cards have the same resolution, lower than that of the input signal I(P n ).

[0132] In one embodiment, the FMi maps are transformed to provide a second set of transformed FMSi feature maps. In this embodiment, the feature vectors are preferably extracted from the transformed maps of the second group, and not directly from the maps of the first group. The maps of the second group are not coded; they are used solely for constructing the feature vectors.

[0133] According to one embodiment, FMi cards are initialized with predefined constant values.

[0134] According to another embodiment, the feature maps are initialized by a set of random real values.

[0135] The FMi feature maps of the first group are subsequently updated, or refined, in step E23, by the encoder's MAJ update module during its training. In step E24, the J zones ZoL of the FMi maps of the first group are determined and encoded by the encoder's FMC module. During the construction phase, this operation is a coding simulation. During the encoding phase, this operation is the actual encoding, and the encoded values ​​constitute the B2 stream. The simulation can be identical to the actual encoding, but it can also be different (for example, simplified). For this encoding, a technique for predicting a feature map value based on its neighborhood is used, as will be described, for example, with reference to Figure 9. In one embodiment, the structure and Ob parameters of the prediction neural network are initialized, for example, during the first iteration of this step.These parameters are subsequently updated, or refined, during the construction phase, in subsequent iterations of the process.

[0136] In one embodiment, the J zone ZoL J FM! cards are coded in the indexing order of the J zones and in a predefined order of associated characteristic cards (i.e., FMjb FM^, FM|, FM4, FM^,...) and the values ​​of each card, for example, lexicographical. Each ZoL zone J FMi cards undergo entropic coding.

[0137] Entropic coding of all ZoL areas of all FMi maps produces a compressed B2 stream whose throughput is subsequently measured during an E29 step.

[0138] In an E25 step, according to one embodiment, the M cards of the first group FMi are transformed by the SE module to generate cards of the second group FMSi at the resolution of the images of the input sequence.

[0139] According to one embodiment, M FMSi cards are generated.

[0140] According to one embodiment, each FMi card is transformed into an FMSi card.

[0141] In one embodiment, at least one FMi map has a lower resolution than the images in the sequence to be encoded, and the transformation operation includes oversampling so that the transformed FMSi map has the same number of samples as the images in the sequence. Oversampling consists of adding values ​​to the FMSi maps to achieve the resolution of the input sequence images. It can be simple (by nearest neighbor replication) or involve interpolation (linear, polynomial, filtered, etc.).

[0142] During step E26, values ​​are extracted by the XTR module from the transformed FM! or possibly FMSi cards. This extraction is performed based on the coordinates (x n , y n) of sample P n of the input signal. It can also be implemented depending on the resolution of the card in question. According to one embodiment, the vector Z n This characteristic results directly from this extraction.

[0143] The samples to be coded are, for example, processed in sequential order, from n=l to n=N.

[0144] According to one embodiment, during step E27, the vector Z n The characteristic is constructed by the TT module from the values ​​extracted from the FM! or FMSi maps for each sample P n coordinates (x n , y n ) of the input signal. The processing may involve quantizing the values ​​extracted from the FM cards or the Z vector n constituted, if necessary. The processing may include other operations, such as filtering, scaling, the application of any function, preferably monotonic, etc.

[0145] In one embodiment, Z n is an M-tuple (zi, Z2, ..., ZM), consisting of the values ​​of the FMi or FMSi cards located at the coordinates (x n , y n ) of the current pixel P n as will be illustrated in support of figure 6.

[0146] In one embodiment, Z n is an M-tuple constructed from values ​​taken from FMi maps at coordinates that may differ between maps. For example, if the FMi maps have different resolutions because they were undersampled, the coordinates are adjusted (by scaling) to match the resolution of each map.

[0147] During step E28, the vector Z n is processed by the NNSYN synthesis neural network to generate as output a vector representative of the sample P n to code, according to one embodiment, the triplet (R, G, B) of the sample P' n(sample P) n (encoded then decoded).

[0148] The structure and Wk parameters of the synthesis neural network are initialized, for example, during the first iteration of this step. These parameters are subsequently updated, or refined, during the construction phase, in later iterations of the process.

[0149] According to one embodiment, the parameters of the synthesis neural network and / or the prediction neural network are initialized by predefined values ​​known to give a satisfactory result (for example, following training on a corpus of images).

[0150] According to another embodiment, the parameters of the synthesis neural network and / or the prediction neural network are initialized by a set of random values.

[0151] During step E29, the Wk parameters of the NNSYN synthesis neural network and the Ob parameters of the ARM prediction neural network, if one exists, are quantized and encoded. During the construction phase, this operation is a coding simulation. During the encoding phase, this operation is the actual encoding, and the encoded values ​​constitute the B1 stream. The simulation can be identical to the actual encoding, but it can also be different (for example, simplified). Any known technique can be used for this purpose, for example, the neural network coding standard proposed by MPEG-7 Part 17, also called NNR (Neural Network Representation). Note that in this case, the amount of degradation that the encoding introduces to the weights Wket (optionally Ob) must be chosen.

[0152] During an E30 step, a performance measure is evaluated.

[0153] To this end, the coding simulation rates associated with the feature maps of the first group (simulation of the B2 flow by coding the J FM J zones of the FM maps,) and optionally with the parameters of the neural network(s) (simulation of the B1 flow by coding the Wket parameters optionally Ob) are measured.

[0154] In one embodiment, the cost function is of the rate-distortion type, denoted (D+L*R), where D, for example, is the root mean square error measured between the input signal and the decoded signal (or the error measured on a subset of the signal samples). In another example, D is calculated from a perceptual function such as SSIM (for Structural SIMilarity) or MSSSIM (for Multi-scale Structural SIMilarity). In one embodiment, R is the simulated rate of stream B1; in another embodiment, R is the total rate used to encode this image, i.e., the sum of the simulated rates of B1 and B2. L is a parameter that controls the rate-distortion trade-off. Other cost functions are possible.

[0155] As long as the cost function has not reached its minimum, or a maximum number of iterations of the cost function minimization algorithm has not been reached, the performance measurement is not satisfactory, and the process is repeated from step E23. This minimization can be performed by a mechanism known as gradient descent with parameter updates during step E23 for the feature map values ​​and E26 / E27 for the network(s) parameters.

[0156] During an EF step, if the cost function has reached its minimum, or if a maximum number of iterations of the cost function minimization algorithm has been reached, training stops. If a coded version corresponding to the last simulation of the synthesis neural network parameters Wk and the feature maps FMj is available, the streams B1 and B2 can be constructed from it. In another embodiment, the actual coding of the updated synthesis neural network parameters Wk and the values ​​of the feature maps FMj and optionally the prediction neural network Ob is performed at this step to produce the encoded parameters Wck (optionally Ob) and FMc, which constitute the streams B1 and B2. Furthermore, at this step, the segmentation Sg is encoded to produce an encoded segmentation of the J masks MZoc J which constitutes the B3 flow.

[0157] The B2 stream comprises sub-streams corresponding to each area of ​​the feature maps. During the encoding of each latent area, it is possible, in a preferred embodiment, to place information identifying the location of the sub-streams within the overall stream at an identifiable point in the stream. This location information can be: a stream pointer (which indicates the address of the beginning of each sub-stream in the overall stream), or a marker (a series of bits otherwise prohibited, which allows traversing the stream to find the beginning of each sub-stream), or any other means of identifying a sub-part of an encoded stream.

[0158] The B1 and B2 streams can be multiplexed and / or concatenated to produce a final stream. In one embodiment, the B3 stream of the coded segmentation and the B1 stream of the coded parameters of the neural network(s) are stored or transmitted before the B2 stream, so that they can be decoded before the B2 stream.

[0159] Figure 6 illustrates a coding method used in one embodiment of the invention.

[0160] In this embodiment, 4 FM! cards are generated. In a preferred mode, there are 7.

[0161] The first FMi card has the same resolution as the I(P) image n ), and therefore includes WxH variables, where W represents the image width in pixels, and H its height. The second FM2 map has half the resolution (in each dimension) of the FM1 map. Each additional map has half the resolution of the previous map. This structure reduces the number of variables in the feature maps, which simplifies coding and learning while minimizing coding costs.

[0162] The FM2 map is oversampled by the SE module by a factor of 2 in each dimension, according to a procedure illustrated in Figure 6. The FM3 map is oversampled by a factor of 4 in each dimension, and the FM4 map by a factor of 8 in each dimension. The FMi map is not affected by the oversampling (FMSi=FMi). The resulting FMSi maps have the same resolution as the I(P) image. n ), and therefore each have WxH values, where W represents the width of the image in pixels and H its height (N=WxH).

[0163] Other types of structure are possible, for example one can use a different reduction rate than half between cards (a quarter, or a third, etc.).

[0164] In this embodiment, the vector Z n is a 4-tuple (Z1...Z4) consisting of the values ​​extracted from the FMSi maps located at the coordinates (x n , y n ) of the current pixel P n The vector Z nconsisting of the extracted (quantified) values ​​from the FMSi maps is processed by the NNSYN synthesis neural network to generate a second output vector; in this example, the output vector is the triplet (R, G, B) of pixel P'. n encoded then decoded. The triplet is inserted into the decoded image I(P' n ) at positions (x n , y n ) of the color components (R', G', B').

[0165] In another embodiment, not shown, the vector Z n is extracted directly from the FMi layers, with positions recalculated according to map size, then the extracted values ​​are possibly processed and quantified after extraction.

[0166] Figure 7 is a flowchart representing an example of a decoding process for a Zod zone. Jwhich can be implemented by the DEC decoding device in Figure 2 when the NNSYN' synthesis neural network is, for example, the one shown in Figure 3.

[0167] During an F20 step, the B1 flow, a part of the B2 flow (that corresponding to the Zod area characteristic maps) J ) and the B3 stream are extracted from the encoded stream. They contain, respectively, Wck parameters, optionally Ocb parameters, an Sgc segmentation, and FMc encoded representations of the M maps of the first group representing the area to be decoded Zod J .

[0168] During an F21 step, an MZo segmentation mask 1 is generated by decoding the values ​​of the encoded segmentation information, or if this information is not encoded, obtained from decoder parameters (size or known position of slices or tiles, etc.).

[0169] During step F22, the M FMdJ maps are generated by decoding the FMcJ values. This decoding uses the MZo mask. 1 and a technique for predicting a feature map value by its neighborhood, as will be described, for example, in support of Figure 10. In one embodiment, the FMdJ maps are decoded in an order, for example lexicographically, (FMd^ . FMd J 2,... FMd J M ) and the values ​​of each card in a predefined order, for example in a row then in a column, taking into account the MZcL mask

[0170] According to embodiments such as those described for the encoder: The FMdJ cards have the same resolution as the Zod signal J (Pdn) to be reconstituted, that is to say that they include N=WxH values.

[0171] The FMdJ cards have a resolution equal to or lower than that of the Zod signal. J (Pdn) to be reconstructed.

[0172] Several FMdJ cards have the same resolution, lower than the signal resolution.

[0173] In step F23, according to one embodiment, the M maps of the first group FMdJ are transformed by the SE' module to generate maps of the second group FMS'J at the resolution of the input images. This step is similar to step E25, which was described for the encoder in support of Figure 5, and the embodiments apply. In particular:

[0174] According to one embodiment, M FMS'Jsont cards are generated.

[0175] According to one embodiment, each FMdJ card is transformed into an FMS'J card.

[0176] In one embodiment, at least one FMdJ map has a lower resolution than the images of the image to be encoded, and the transformation operation includes oversampling so that the transformed FMS'J map has the same number of samples as the input image. Oversampling consists of adding values ​​to the FMS'J maps to achieve the resolution of the input image. It can be simple (by nearest neighbor replication) or involve interpolation (linear, polynomial, by fracturing, etc.).

[0177] The transformation may optionally include inverse quantization of the extracted values, if necessary. However, inverse quantization is not mandatory.

[0178] During step F24, values ​​are extracted by the XTR' module from the transformed FMdJ or possibly FMS'J maps. This extraction is performed based on the coordinates (x n , y n ) of a sample Pn of the input signal. It can also be implemented depending on the resolution of the card in question. This step is similar to step E26, which was described for the encoder in support of Figure 5, and the embodiments apply. In particular:

[0179] In one embodiment, Zd n is an M-tuple (zi, Z2, ..., ZM), consisting of the values ​​of the FMdJ or FMS'J maps located at the coordinates (x n , y n ) of a current pixel Pd n as will be illustrated in support of figure 8.

[0180] The samples to be decoded are, for example, processed in sequential order, relative to the MZo mask. J According to one embodiment, during a step F35, a vector Zd n is constructed by the TT' module from the values ​​extracted from the FMdJ maps of the first group or the FMS'J maps of the second group, for each sample Pd n coordinates (x n , y n) to decode, based on the coordinates (x n , y n This step is similar to step E27, which was described for the encoder in support of Figure 5, and the described embodiments apply. The extraction may involve inverse quantization of the extracted values ​​or of the vector Zd n constituted, if necessary.

[0181] During step F26, the Wdk parameters of the NNSYN' synthesis neural network and, optionally, the Odb parameters of the prediction neural network are generated by decoding the Wck and Ocb values ​​of the B1 stream. Any known decoding technique corresponding to the encoding technique used by the encoder can be used for this purpose. The NNSYN' synthesis neural network is similar to the NNSYN synthesis network; that is, it has the same structure and parameters, except for the encoding, which can be performed with or without loss. Similarly, the ARM' prediction neural network, if used to decode feature maps, is similar to the ARM prediction network; that is, it has the same structure and parameters, except for the encoding, which can be performed with or without loss.

[0182] In one embodiment, stream B1 is decoded before streams B2 and B3, so that the synthesis neural network NNSYN' and possibly the prediction neural network ARM' are available before decoding the samples. Similarly, stream B3 is decoded before stream B2 so that the mask of the region to be decoded is available before decoding the samples.

[0183] During step F27, the vector Zd n is processed by the NNSYN synthesis neural network 1 to generate as output a second vector representing the sample Pd n to decode, according to one embodiment, a triplet that is injected into the image of the decoded zone Zod J (Pd n ) at positions (x n , y n ) of the color components (Rd, Gd, Bd). This step is similar to step E28 which was described for the encoder in support of Figure 5.

[0184] When all the signal samples have been processed, the decoded signal corresponds to the Zod area J (Pdn) is available.

[0185] The decoding of one area of ​​the image has been described. Decoding a second area can be performed successively or in parallel, since the areas are encoded independently of each other. If the areas are decoded in parallel, it is necessary to have information indicating the precise position of the sub-streams (latent data) corresponding to the different areas (for example, sub-stream pointers). On the other hand, if the areas are decoded sequentially, one can move from one sub-stream to another to decode the areas one after the other, without any specific indication.

[0186] Figure 8 illustrates a method for decoding a Zod zone. J used in one embodiment of the invention.

[0187] In this embodiment, there are 4 FMdJ decoded cards. In a preferred mode, there are 7.

[0188] In this embodiment, the decoded FMdJ maps are representative only of the Zod zone J In other words, only data from the ZoL area J (corresponding in the latent domain to the Zod zone) J ) have been decoded.

[0189] In this embodiment, the first FMd card has the same resolution as image I, and therefore has variable WxH, where W represents the width of the image in pixels, and H its height. The second FMd card J The second map has half the resolution (in each dimension) of the FMd^ map. Each additional map has half the resolution of the previous map. This structure reduces the number of variables in the feature maps, which facilitates decoding while minimizing the encoding cost.

[0190] The FMdJ map is oversampled by a factor of 2 in each dimension, using any oversampling method available to a person skilled in the art. The FMd map J 3est is oversampled by a factor of 4 in each dimension, and the FMd^ map by a factor of 8 in each dimension.

[0191] The FMS'J cards have the same resolution as the image to be decoded, and therefore include WxH values, where W represents the width of the image in pixels, and H its height.

[0192] In this embodiment, the vector Zd n is a 4-tuple (Z1...Z4) consisting of the values ​​of the FMS'J cards located at the coordinates (x n , y n ) of the current pixel Pd n The vector Zd n is optionally dequantized and then processed by the NNSYN' synthesis neural network to generate the output triplet (R, G, B) representative of the Pd sample n to be decoded. The triplet (R, G, B) is inserted into the decoded image I(Pd n) at the coordinates (x n , y n ) in the color components (Rd, Gd, Bd). Figure 9 is a flowchart representing a method for encoding feature maps that can be implemented by the encoding device of Figure 1 and by the encoding process of Figure 5.

[0193] These steps are substeps of step E24 described previously in Figure 5. Their purpose is to encode a current value V n of a ZoL zone J of an FM characteristic map! of the first group being processed using values ​​from a neighborhood of the current value V n .

[0194] During a substep E241, a neighborhood vector C n is established, including values ​​close to the value V nAs will be illustrated later with support from Figures 11 to 16, these neighboring values ​​can be located in the same map and / or in a different map from the plurality M of FM maps! This neighborhood vector consists of a number C of values ​​(for example, C=4, C=8, C=9 or C=11 in the examples of Figures 11 to 16), or data, corresponding to neighborhood values ​​of the current value V n when these neighborhood values ​​belong to the ZoL zone J or alternative values ​​if necessary. These values ​​must be known to the encoder and decoder; therefore, they must be in a causal neighborhood of the value V. n .

[0195] According to a first embodiment, these values ​​are used to determine the context of an entropy encoder to encode the current value during an E244 step. This encoder can be a CAB AC (Context-Adaptive Binary Arithmetic Coding) encoder. This type of encoder is well known to those skilled in the art. It is notably used in the H.265 / HEVC video compression standard. It is an arithmetic encoder with lossless compression. It decomposes all non-binary symbols into binary symbols. Then, for each bit, the encoder selects the most suitable probability model and uses a context to optimize the probability estimation. This context can be defined by information from neighboring elements. Adaptive or non-adaptive arithmetic coding is then applied to compress the resulting data. As is known to those skilled in the art, there are several ways to use the neighborhood vector to generate context information.For example, one can count the number of neighboring values ​​that are not zero and associate a context with each number. Alternatively, one can perform comparisons between several neighboring values ​​and associate a given context with a configuration of order between the neighboring values, for example by ranking the neighboring values ​​in ascending order and associating a context with each possible order.

[0196] In a second embodiment, the neighborhood is used to predict, during step E242, the current value from an autoregressive model. Recall that an autoregressive model predicts a sample from a series based on its past values. In this embodiment, the past values ​​are constituted by the context, and the difference between the predicted variable and the actual value is quantified and then encoded entropically during step E244.

[0197] In a third embodiment, as illustrated in Figures 11 to 16, an ARM predictive neural network is used in step E243 to predict the statistical characteristics of the variable to be coded. The neighborhood vector is given as input to the ARM network, which provides an output prediction of the current value. In one embodiment, the ARM network behaves as a function f v as described in Figure 4, which provides a set of statistical parameters (mean, variance, median, etc.) used to entropically code the current value. The role of this ARM module is to predict the current value as accurately as possible, for all values ​​of V n to code the ZoL zone JFM maps are used to reduce the throughput required for encoding feature maps. In another embodiment, the ARM predictive neural network is used to generate the expected probability (pr) of the possible value of the current sample. Entropy coding is then adapted to this probability (as is known for Huffman or arithmetic entropy coding).

[0198] At the end of the process, the current coded value Vc n The FM card! is being processed and is coded.

[0199] Figure 10 is a flowchart representing a method for decoding feature cards that can be implemented by the decoding device in Figure 2 and by the decoding process in Figure 7.

[0200] These steps are substeps of step F22 described earlier, supported by Figure 7. Their purpose is to decode a current value Vd nfrom a point on an FMdJ feature map of the first group being processed using neighborhood values.

[0201] During a substep F221, a neighborhood vector Cd n is established, including values ​​close to the value Vd n This step is similar to step E241 described earlier, and the same embodiments apply. This neighborhood vector consists of a number C of values, or data, corresponding to neighborhood values ​​(for example, C=4, C=8, C=9, or C=11 in the examples in Figures 11 to 16) located in the same map and / or in a different map from the plurality M of maps FMdJ. These values ​​lie in a causal neighborhood of the value Vd nare known to the decoder. According to a first embodiment, these values ​​are used to determine the context of an entropic decoder for decoding the current value during an F224 step. This decoding is similar to that used in the encoder, for example, CABAC. The use of the neighborhood to produce context information is similar to that chosen in the encoder. For example, one can count the number of non-zero neighboring values ​​and associate a context with each number. Alternatively, one can perform comparisons between several neighboring values ​​and associate a given context with an ordering configuration among the neighboring values, for example, by sorting the neighboring values ​​in ascending order and associating a context with each possible order.

[0202] In a second embodiment, the neighborhood is used to predict the current value from an autoregressive model during step F222. In this mode, the past values ​​are constituted by the context, and the difference between the predicted variable and the actual value is decoded and then entropically dequantized during step F224.

[0203] In a third embodiment, as illustrated in Figures 11 to 16, a predictive neural network ARM' is used in step F223 to predict the statistical characteristics of the variable to be decoded. The neighborhood vector is applied as input to the ARM' network to provide an output prediction of the current value. In one embodiment, the ARM network behaves as a function f vas described in support of Figure 4, defined by a set of statistical parameters (mean, variance, median, etc.) used to entropically decode the current value. In another embodiment, the ARM' prediction neural network is used to produce the expected probability (pr) of the possible value of the current sample. Entropic decoding is adapted to this probability (as is known for Huffman or arithmetic entropic coding). The ARM' network is identical to the ARM network if its coding has been lossless. At the end of the process, the current decoded value Vd n The FMdJ card currently being processed is decoded.

[0204] Figures 11 and 12 illustrate a method for encoding or decoding feature maps according to one of the embodiments

[0205] In this illustration, the encoding (rep. decoding) of the current value Vn (resp. Vd n ) located at a current point with coordinates (x n , y n ) in the characteristic zone ZoL J The current feature map FM! (resp. FMdJ) uses the contextual information of this current map. This contextual information is represented by a neighborhood of the current point V n In the example presented here, the neighborhood used includes 8 points belonging to a causal neighborhood of the current point associated with a 3x3 square mask.

[0206] This contextual information is, however, conditioned by the characteristic zone ZoL J in the current map. More specifically, the contextual information depends on the location of the current point relative to the ZoL boundary. J This location is obtained by evaluating a proximity criterion Cp(V n ) (resp. Cp(Vd n)) of the current point with respect to a boundary of the characteristic zone. In the example described here, the proximity criterion Cp(V n ) (resp. Cp(Vdn)) is the Euclidean distance between the current point and the zone boundary ZoL J Alternatively, the proximity criterion Cp(V n ) (resp. Cp(Vd n )) can be a Manhattan, Minkowski, or Chebyshev distance. Alternatively, the distance can simply be measured in pixels, for example, by counting the number of pixels along the two directions of a coordinate system associated with the current map between the current point and the boundary of the ZoL area. J In yet another variant, the proximity criterion Cp(V n ) (resp. Cp(Vdn)) may also be a quasi-distance or a gap or any other relevant measure indicating the proximity of the current point to the boundary of the ZoL zone J .

[0207] Two cases must then be distinguished:

[0208] In the first case, the proximity criterion Cp(V n ) (resp. Cp(Vd n is greater than a threshold S (e.g., predefined) and the values ​​of the neighborhood points in the current map FM! (resp. FMdJ) are used to encode (resp. decode) the current value V n (resp. Vd n ).

[0209] In the second case, the proximity criterion Cp(V n ) (resp. Cp(Vd n is less than this threshold S. In this case, the neighborhood vector C n (resp. Cdn) is defined component by component. The value of a component of the neighborhood vector C n (resp. Cd n ) is equal to the value of the associated neighborhood point in the current map if that point belongs to the ZoL zone J and to a replacement value otherwise.

[0210] Figure 11 illustrates the first case while Figure 12 illustrates the second case.

[0211] In Figure 12, the values ​​of the points shown in grey are part of the ZoL zone J and are used to determine components of the neighborhood vector C n The values ​​of the points shown in white do not belong to the ZoL zone. J but in another ZoL zone x and cannot be used to define the neighborhood vector C n They are replaced by a replacement value Rp. In one embodiment, such as that described in Figure 12, the replacement value is a constant, for example equal to the average value of the points in the ZoL zone. J in the current card or at a predetermined constant value.

[0212] In another embodiment shown in Figure 13, the replacement value is calculated based on the value of the neighborhood points belonging to the ZoL zone J .

[0213] In another embodiment, a set of replacement values ​​is calculated based on the values ​​of the neighborhood points belonging to the ZoL zone. J A neural network can be implemented to calculate a given number of replacement values ​​based on points in the neighborhood of the ZoL area. J .

[0214] In the previous embodiments, the definition of the neighborhood of the current point and the dimension of the neighborhood vector C n are independent of the proximity criterion Cp(V n ) (resp. Cp(Vdn)). In another embodiment, this form of the neighborhood and / or the dimension of the neighborhood vector C n may depend on the proximity criterion Cp(V n ) (resp. Cp(Vd n )).

[0215] Figure 14 presents an example where the definition of the neighborhood is independent of the proximity criterion Cp(V n ) (resp. Cp(Vd n)) but where is the dimension of the neighborhood vector C n depends on this proximity criterion Cp(V n ) (resp. Cp(Vd n )).

[0216] Thus, when the proximity criterion Cp(V n ) (resp. Cp(Vd n )) is greater than the threshold S, the neighborhood vector C n is determined as explained with reference to Figure 11.

[0217] However, when the proximity criterion Cp(V n ) (resp. Cp(Vd n )) is less than the threshold S, the neighborhood vector C n is determined as follows: the number of points Np in the causal neighborhood of the current point belonging to the characteristic zone ZoL J is calculated and the neighborhood vector V n is defined as a vector of dimension Np, and the value of each point in the neighborhood of the current point belonging to the ZoL zone Jis associated, according to a predefined order for example, with a different component of the neighborhood vector C n .

[0218] Furthermore, in the preceding embodiments, the contextual information represented by the neighborhood of the current point is independent of the proximity criterion Cp n Alternatively, and as shown in Figure 15, the definition of the neighborhood of the current point can depend on the proximity criterion Cp n .

[0219] Two cases must again be distinguished: In the first case, the proximity criterion Cp(V n ) (resp. Cp(Vd n is greater than the threshold S and the neighborhood values ​​of the current card FM! (resp. FMdJ) belonging to a first neighborhood (identical to that defined previously in relation to Figure 11) are used to encode (resp. decode) the current value Vn (resp. Vd n ).

[0220] In the second case, the proximity criterion Cp(Vn ) (resp. Cp(Vd n is less than this threshold S. In this case, a second neighborhood of the current point is used. In the example shown in Figure 15, the neighborhood used is a causal neighborhood comprising 10 points included in a 4x3 mask of the current point. The neighborhood vector C n (resp. Cd n The current point's vector is then defined component by component. The value of a component of the neighborhood vector C n (resp. Cd n ) is equal to the value of the associated point in the second neighborhood if that point belongs to the ZoL zone J and to a replacement value otherwise.

[0221] Once determined, the neighborhood vector C nis used to encode (resp. decode) the current value in one of the embodiments described previously in support of Figure 9 (resp. Figure 10). Figure 16 illustrates another method of encoding or decoding feature maps according to one of the embodiments.

[0222] In this illustration, the encoding (resp. decoding) of the current value V n (resp. Vd n ) located in coordinates (x n , y n ) in the current feature map FMi (i.e., the one being encoded or simulated encoding) (resp. FMd^ - i.e., the one being decoded) uses the contextual information from its own map and the previous map FM2 (resp. FMd2). The values ​​located at the coordinates (x n , y n -l), (x n , y n -2), (x n -l, y n -l), (x n -l, y n ), (x n-l, yn+1), (Xn+l, y n -2), (x n +l, yn-l), (x n +l, y n ), (x n +l, y n +l) in the FMi layer (resp. FMdJ , and the value located at the coordinates (x n , y n ) in the FM2 layer (resp.

[0223] FMd2) are used to determine the neighborhood to encode (resp decode) the current value.

[0224] This contextual information is conditioned by the ZoL zone J in the current map and in the previous map. Contextual information depends on the location of the current value V n (resp. Vdn) in relation to the ZoL zone boundary J in the current map and in the previous map. The neighborhood vector is determined differently depending on the evaluation of the proximity criterion Cp(Vn) (resp. Cp(Vdn)). If the proximity criterion Cp(V n ) (resp. Cp(Vd nIf the current value is greater than a threshold S, the neighborhood values ​​of the current and previous maps are used to encode (resp. decode) the current value. Otherwise, the neighborhood vector Cn (resp. Cd) is used. n ) of the value V n (resp. Vd n ) is defined component by component. The value of a component of the neighborhood vector C n (resp. Cd n ) is equal to the value of the associated point of the current map or of the previous map if that point belongs to the ZoLj zone and to a replacement value otherwise.

[0225] Figure 16 illustrates this latter case with 9 points from a causal neighborhood included in a 3x3 mask in the current map and one pixel in the previous map. The values ​​shown in gray are part of the ZoL area J and are used to determine components of the neighborhood vector C nThe values ​​shown in white do not belong to the ZoL zone. J but in the ZoL area x and cannot be used to define the neighborhood vector C n , they are replaced by a replacement value Rp.

[0226] Once determined, the neighborhood vector C n is used to encode (resp. decode) the current value in one of the embodiments described previously in support of Figure 9 (resp. Figure 10).

[0227] In the embodiments illustrated in Figures 11 to 16, the neighborhood vector C n is extracted by a CTX module (resp. CTX') from the FMC (resp. FMD) coding module, then it is applied to the input of the ARM (resp. ARM') prediction neural network, used to predict the statistical characteristics (p,o) or the probability (pr) of the value to be coded (resp. decoded) by the CE (resp. DE) entropy coder.

[0228] It should also be noted that the invention is not limited to the embodiments described above. Indeed, it will be apparent to those skilled in the art that various modifications can be made to the embodiments described above, in light of the information just disclosed to them.

[0229] For example, NNSYN / NNSYN' synthetic neural networks can be recurrent neural networks.

[0230] In another example, synthetic neural networks can consist of one or more convolutional neural networks, followed by a machine learning program (MLP), and then followed by one or more convolutional neural networks, each convolutional neural network being associated with a convolutional kernel of predefined size, for example, 3x3. In these examples, obtaining the Z vectors n / Zd nis adapted to the topology of NNSYN / NNSYN' synthesis neural networks, meaning that several values ​​are extracted from at least one FM feature map. In other words, Z n is an L-tuple with L>M, constructed from values ​​taken from the FM! maps at coordinates close to those of the current pixel, for example on the support of the kernel of the convolution filter of one of the convolutional neural networks.

[0231] In the detailed presentation of the invention given above, the terms used shall not be interpreted as limiting the invention to the embodiments set forth in this description, but shall be interpreted as including all equivalents which can be foreseen by a person skilled in the art by applying their general knowledge to the implementation of the teaching which has just been disclosed to them.

[0232] According to a particular embodiment, the invention relates to a method for encoding a signal comprising at least two zones, comprising the following steps for a zone to be encoded of said signal comprising a plurality of samples to be encoded: the construction of a group of latent value maps representative of the signal, obtaining a segmentation of said group of latent value maps into at least two characteristic zones, one of the at least two characteristic zones, called the associated characteristic zone, being associated with said zone to be encoded, for at least one sample, called the current sample, of the zone to be encoded, associated with a position in the signal to be encoded and with a point of the associated characteristic zone, called the associated point:

[0233] • the construction of a characteristic vector from said latent value maps of said group, as a function of said position of said current sample,

[0234] • the application of said characteristic vector to the input of an artificial neural network, called a synthetic neural network, defined by a set of parameters, to obtain, at the output of said synthetic neural network, a vector representing a decoded value of the current sample,

[0235] • the encoding of a value, called the current value of said associated point, said encoding of the current value comprising the evaluation of a proximity criterion of said associated point with respect to a boundary of said characteristic area, the construction of a neighborhood vector of said associated point from a neighborhood of said associated point in said latent value maps of said group and the evaluation of said proximity criterion, and an entropic encoding of said current value as a function of said neighborhood vector, and

[0236] • the updating of at least one value of one of said latent value maps of said group and / or of at least one parameter of said synthetic neural network, according to a coding performance measure, the coding of a group of latent value maps representative of said area to be coded of said signal, and the coding of said set of parameters of said synthetic neural network.

[0237] According to a particular embodiment, the invention also relates to a method for decoding a region, called the decoding region, of a signal comprising at least two regions, said decoding region comprising a plurality of samples to be decoded, said decoding method comprising the following steps: decoding a group of latent value maps representative of said decoding region of said signal, said decoding comprising

[0238] • the determination in said latent value maps of said group of a characteristic zone corresponding to said zone to be decoded of said signal,

[0239] • for at least one value, called current value, of a point in said characteristic zone, the evaluation of a proximity criterion of said point with respect to a boundary of said characteristic zone, the construction of a neighborhood vector of said point from a neighborhood of said point in said latent value maps of said group and the evaluation of said proximity criterion, and an entropic decoding of said current value as a function of said neighborhood vector, the decoding of a set of parameters representative of a neural network, called a synthesis neural network, and the processing of said group of latent value maps by the synthesis neural network.The decoding method (and symmetrically the encoding method) may further include one or more of the following optional features, taken individually or in any technically feasible combination: the neighborhood of said point is independent of said evaluation of the proximity criterion. According to another feature, the neighborhood of said point is selected based on said evaluation of the proximity criterion during the construction of the neighborhood vector; a component of the neighborhood vector is associated with a point of said neighborhood, the value of said component being equal to the value of said associated point if said associated point is a point of said characteristic zone and to a replacement value otherwise; the replacement value is dependent on the points of said characteristic zone or on the points associated with said neighborhood belonging to said characteristic zone.The replacement value does not depend on the points of said characteristic zone. The neighborhood of said point is selected based on said point. The construction of the neighborhood vector includes: identifying the points of said neighborhood belonging to said characteristic zone, called identified points; initializing the number of components of said neighborhood vector to the number of identified points; and assigning the value of each identified point to a different component of said neighborhood vector. The proximity criterion of said point with respect to a boundary of said characteristic zone is a distance. The decoding process further includes a step of obtaining a segmentation of said signal into at least two zones, and said characteristic zone is determined from said segmentation of said signal.The decoding process further includes a step of obtaining a mask of the area to be decoded in the signal, and the characteristic area is determined from this mask. The neighborhood vector is used to determine a context for the entropic decoding of the current value. The entropic decoding of the current value includes a step of predicting the current value based on the neighborhood vector. The decoding of the current value from one of the latent value maps includes: - decoding a set of parameters representative of a neural network, called the prediction neural network, and processing the neighborhood vector by the prediction neural network to provide a prediction of the current value.

Claims

1. Demands 1. Method for encoding a signal (I(P n )) comprising at least two zones, comprising the following steps for one zone (Zo 1 ) comprising a plurality of samples (P n ) to be coded: the construction (E22) of at least one feature map representative of said area, for at least one sample of said area ( / o'), called the current sample (P n ), associated with a point in said at least one feature map, the encoding of a data associated with said point, called the current value (V n ), including: o the evaluation (F221) of a proximity criterion of said point with respect to a boundary associated with said zone, o the construction (F221) of a neighborhood vector (C n ) of said point according to said proximity criterion, and an entropy coding (F223, F224) of said current value from said neighborhood vector (C n), the update (E23, E26, E27) of at least one value of said feature map, based on a coding performance measure, the coding of said feature map (FMcJ) representative of said area.

2. A method for decoding a data stream representative of a signal segmented into zones, said decoding method comprising the following steps, for a zone (Zod J ): the decoding of at least one characteristic map (FMdJ) representative of said zone (Zod J ), the decoding including, for at least one point of said map (FMdj): • the evaluation of a proximity criterion for said point in relation to a boundary associated with said area, • the construction (F221) of a vector (C n ), said neighborhood vector of said point, according to said proximity criterion, and • an entropic decoding (F224) of a data point associated with the point, called the current value (V n), from said neighborhood vector (C n ), the application of a neural network (NNSYN') to said at least one decoded feature map.

3. Decoding method according to claim 2 further comprising a decoding step (F26) of a set of parameters (Wdk) representative of said synthetic neural network (NNSYN').

4. A decoding method according to claim 2 further comprising a decoding step (F26) of segmentation information to determine (F21), in said at least one feature map (FMdJ), a feature area (ZoL) J ) corresponding to said zone (Zod J ).

5. Decoding method according to claim 2 wherein the neighborhood of said point is independent of said evaluation of the proximity criterion.

6. A decoding method according to claim 2, wherein the neighborhood of said point is selected based on said evaluation of the proximity criterion 7. A decoding method according to any one of claims 4 to 6, wherein, during the construction of the neighborhood vector, a component of the neighborhood vector is associated with a point of said neighborhood, the value of said component being equal to the value of said associated point if said associated point is a point of said characteristic zone (ZoL) J ), and to a replacement value, otherwise.

8. A decoding method according to the preceding claim, wherein said replacement value is dependent on the points of said characteristic zone (ZoL J ) or points associated with said neighborhood belonging to said characteristic zone (ZoL J ).

9. A decoding method according to claim 7, wherein said replacement value does not depend on the points of said characteristic zone (ZoL J ).

10. A decoding method according to any one of claims 4 to 9, wherein the construction of the neighborhood vector (C n ) includes: the identification of the points of said neighborhood belonging to said characteristic zone (ZoL J ), said identified points, the initialization of the number of components of said neighborhood vector (C n ) to the number of points identified, and assigning the value of each identified point to a different component of said neighborhood vector (C n ).

11. Decoding method according to any one of claims 2 to 10 wherein said proximity criterion of said point with respect to a boundary of an area is a distance.

12. A decoding method according to any one of claims 2 to 11, wherein said neighborhood vector (C n ) is used to determine a context for the entropic decoding of said current value (V n ).

13. A decoding method according to any one of claims 2 to 12, wherein said entropic decoding of said current value (Vn) comprises a step of predicting said current value as a function of said neighborhood vector (C n ).

14. A decoding method according to any one of claims 2 to 13, characterized in that the decoding of said current value (V n ) of one of said feature maps includes: the decoding of a set of parameters (Odk) representative of a neural network, called a predictive neural network (ARM'), and the processing of said neighborhood vector (C n) by the prediction neural network (ARM'), to provide a prediction of said current value (V n ).

15. Encoding device (ENC) of a signal (I(P n )) comprising at least two zones, characterized in that said coding device comprises a processor coupled to a memory in which are stored instructions configured to implement for a zone (Zo 1 ) of said signal comprising a plurality of samples (P n ) to be coded: the construction (E22) of at least one feature map representative of said area, for at least one sample, called the current sample (P n ), of said zone ( / o'), associated with a point in said at least one feature map, the encoding of a data associated with said point, called current value (V n ), including: • the evaluation (F221) of a proximity criterion for said point in relation to a boundary of said zone, • the construction (F221) of a neighborhood vector (C n ) of said point according to said proximity criterion, and • an entropy coding (F223, F224) of said current value as a function of said neighborhood vector (C n ), the update (E23, E26, E27) of at least one value of said feature map, based on a coding performance measure, the coding of said feature map (FMc) representative of said area.

16. A device for decoding a data stream representative of a signal segmented into zones, the device comprising a processor coupled to a memory in which instructions configured to implement the following steps are stored for a zone (Zod). J ): the decoding of at least one characteristic map (FMdJ) representative of said zone (Zod J ), the decoding including, for at least one point of said map (FMdj): • the evaluation of a proximity criterion for said point in relation to a boundary of said zone, • the construction (F221) of a neighborhood vector (C n ) of said point according to said proximity criterion, and • an entropic decoding (F224) of a data point associated with the point, called the current value (V n ), from the neighborhood vector (C n ), the application of a neural network (NNSYN') to said at least one decoded feature map.

17. Computer program comprising instructions for carrying out the steps of an encoding process according to claim 1 or a decoding process according to any one of claims 2 to 14 when said program is executed by a computer.

Images