FR3057131A1 - METHOD FOR ENCODING A DIGITAL IMAGE, DECODING METHOD, DEVICES, TERMINAL AND COMPUTER PROGRAMS - Google Patents

Abstract

A method of encoding an image of a sequence of digital images, said image being divided into a plurality of blocks of pixels processed in a defined order, said method comprising the following steps, implemented for a block, called current block, of the current image, of predetermined dimensions: Prediction (E1) of the values of the current block from at least one previously processed block of a previous or next image, called reference image, Calculation ( E2) of a residue block (R) by subtraction of the predicted values from the original values of the current block; Transformation (E6) of the residual block by application of a transform, the transform belonging to a predetermined list of transforms; Coding (E7) of the transformed residue block and identification information of the applied transform; According to the invention, the method further comprises a step (E3) of obtaining an indicator of variability of the block with respect to the at least one reference image, a step (E4) of obtaining a list of transformed from among a plurality of predetermined lists at least according to the indicator obtained and the transformation step applies a transformed list obtained.

Description

Holder (s): BoCOM, ORANGE Public limited company, TDF Simplified joint stock company.

Extension request (s)

Agent (s): AVOXA.

METHOD FOR ENCODING A DIGITAL IMAGE, METHOD FOR DECODING, DEVICES, TERMINAL AND COMPUTER PROGRAMS THEREOF.

FR 3 057 131 - A1 (57) The invention relates to a method of coding an image of a sequence of digital images, said image being divided into a plurality of blocks of pixels processed in a defined order, said method comprising the following steps, implemented for a block, called the current block, of the current image, of predetermined dimensions:

Prediction (E 1) of the values of the current block from at least one block previously processed from a previous or next image, called the reference image,

Calculation (E2) of a residual block (R) by subtracting the predicted values from the original values of the current block,

Transformation (E6) of the residual block by application of a transform, the transform belonging to a predetermined list of transforms;

Coding (E7) of the transformed residual block and of information identifying the applied transform;

According to the invention, the method further comprises a step (E3) of obtaining an indicator of variability of the block with respect to the at least one reference image, a step (E4) of obtaining a list of transforms among a plurality of predetermined lists at least as a function of the indicator obtained and the transformation step applies a transform of the list obtained.

i

Method for coding a digital image, method for decoding, devices, terminal equipment and associated computer programs

1. Field of the invention

The field of the invention is that of signal compression, in particular of a digital image or of a sequence of digital images, divided into blocks of pixels.

The invention relates more particularly to the signaling of a transform applied to a block of pixels, in a context of transform competition.

The coding / decoding of digital images applies in particular to images from at least one video sequence comprising:

- images from the same camera and successively in time (2D type coding / decoding),

- images from different cameras oriented according to different views (3D type coding / decoding),

- corresponding texture and depth components (3D type coding / decoding),

- etc.

The present invention applies similarly to the coding / decoding of 2D or 3D type images.

The invention can in particular, but not exclusively, be applied to the video coding implemented in current video coders AVC (for “Advanced Video Coding”, in English) and HEVC (for “High Efficiency Video Coding”, in English) and their extensions (MVC, 3D-AVC, MV-HEVC, 3D-HEVC, etc.), and to the corresponding decoding.

2. Presentation of the prior art

We consider a classical compression scheme of a digital image, according to which the image is divided into blocks of pixels. A current block to be coded, which constitutes an initial coding unit, is generally divided into a variable number of sub-blocks according to a predetermined cutting mode.

In relation to Figure 1, we consider a sequence of digital images Ii, I2, Ik, with K non-zero integer. An Ik image is divided into initial coding units or CTU (for “Coding Tree Unit” in English) according to the terminology of the HEVC standard, as specified in the document ISO / IEC 23008-2: 2013 - High efficiency coding and media delivery in heterogeneous environments - Part 2: High efficiency video coding ”, International Organization for Standardization, published in November 2013 or in the ITU-T H.265 specification entitled“ High Efficiency Video Coding ”, published in April 2015.

. Standard coders generally offer regular partitioning, based on square or rectangular blocks, called CU (for “Coding Units”, in English) of fixed size. Partitioning is always done from the initial coding unit, not partitioned, and the final partitioning is calculated and then signaled from this neutral base.

Each CU undergoes an encoding or decoding operation consisting of a series of operations, including in a non-exhaustive manner a prediction, a residue calculation, a transformation, a quantification and an entropy coding. This sequence of operations is known from the prior art and presented in relation to Figure 2.

The first CTU block to be processed is selected as the current block c. For example, this is the first block (in lexicographic order). This block comprises NxN pixels, with N non-zero integer, for example equal to 64 according to the HEVC standard.

During a step E1, a prediction Pr of the original block b is determined. It is a prediction block constructed by known means, typically by motion compensation (block from a previously decoded reference image), or by intra prediction (block constructed from decoded pixels belonging to the image ID). The prediction information linked to Pr is coded in the bit stream TB or compressed file FC. We assume here that there are P possible prediction modes rrii, m _p , with P non-zero integer. For example, the prediction mode chosen for the current block c is the mode m _p . Certain prediction modes are associated with an Intra type prediction, others with an INTER type prediction.

During a step E2, an original residue R is formed, by subtraction R = c-Pr from the prediction Pr from the current block c to the current block c.

In E3, a transform is identified to be applied to the residue R obtained.

The transformation stage plays a crucial role in such a video coding scheme: it is in fact it which concentrates the information before the quantification operation. As a result, a set of residual pixels before encoding is represented on a small number of non-zero frequency coefficients representing the same information. Thus, instead of transmitting a large number of coefficients, only a small number will be necessary to faithfully reconstruct a block of pixels.

In image and video coding, block transforms (4x4, 8x8, etc.), orthogonal or quasi-orthogonal are generally used. The most widely used transforms are based on cosine bases. The DCT is thus present in most standards for image and video. Recently, the HEVC standard has also introduced DST (for “Discrete Sine Transform”, in English) for coding particular residues in the case of blocks of 4x4 size.

In fact, it is approximations of these transforms that are used, the calculations being carried out on integers. In general, the bases of transforms are approximated to the nearest integer, on a given precision (generally of 8 bits).

By way of example, in relation to FIGS. 2A and 2B, the transforms used by the HEVC standard on 4x4 size blocks are presented: These are DCT and DST transforms. The values presented in this table are to be divided by 128 to find the quasiorthonormal transformations.

More recently the concept of transform competition has been introduced. For a given block size and prediction mode, the encoder has the choice between T transformed, with non-zero integer T, generally greater than or equal to 2. In the same way as for partitioning a block, it each applies in turn to the current block, then evaluates them according to a flow-distortion criterion. The transform chosen is the one that obtains the best performance:

The HEVC standard provides, for 4x4 blocks, the choice between a DST type transformation or the absence of a "Transform Skip" transform (that is to say that the residual coefficients do not undergo a transform).

In the publication by A. Arrufat et al, entitled “Rate-distortion optimized transformed competition for intra coding in HEVC”, published in the Proceedings of the IEEE Visual Communication on Image Processing conference, in December 2014, held at La Valletta, Malta. pp.73, the encoder has the choice between multiple transforms. For example, 5 transforms are proposed for 4x4 size blocks and 17 transforms for 8x8 size blocks. Step E3 therefore identifies a transform among those available as a function of the size of the current block.

During a step E4, the residue R is transformed into a transformed residue block, called RT, by the identified transform. It is for example a block type transform (commonly DCT or even DST or even adapted transformations) or a wavelet transform, all known to those skilled in the art and notably implemented in the JPEG / MPEG standards for DCT / DST and JPEG2000 for transforming into wavelets.

In E5, as is known in the state of the art, these coefficients are scanned in a predetermined order so as to constitute a one-dimensional vector RQ [j], where the index j varies from 0 to Nb-1, with Nb integer equal to the number of pixels in block c. The index j is called the frequency of the coefficient RQ [j]. Conventionally, these coefficients are scanned in globally increasing or decreasing order of frequency values, for example according to a zigzag path, which is known from the JPEG fixed image coding standard.

In E6, the transformed residue RT is quantified by conventional quantification means, for example scalar or vector, into a quantized residue block RQ comprising as many coefficients as the residue block RQ contains pixels, for example Nb, with non-zero integer Nb .

During a step E7, the information relating to the coefficients of the residual block RQ is coded by entropy coding, for example according to a Huffman coding technique or arithmetic coding. This information includes at least the amplitude of the coefficients, their sign and a signaling of the transform applied to the block by the encoder. By amplitude is meant here the absolute value of the coefficient. Conventionally, information representative of the fact that the coefficient is not zero can be coded for each coefficient. Then, for each non-zero coefficient, one or more pieces of information relating to the amplitude are coded. The amplitudes coded CA are obtained. We also code the signs of the non-harmful coefficients. In general, they are simply coded by a bit 0 or 1, each value corresponding to a given polarity. Such coding obtains efficient performance because, due to the transformation, the values of the amplitudes to be coded are largely zero.

Concerning the applied transform, in the case of the HEVC standard, one indicates to the decoder by a bit, called transform_skip_flag, the inverse transform to be applied among the two alternatives DST or absence of transform.

In the case of the publication of A. Arrufat et al, it is signaled to the decoder by an indicator approach plus explicit signaling: the encoder signals by an indicator (indicator in position 0) if the transform is a HEVC type transform (according to size, it is a DCT or a DST) or a specific transformation (indicator in position 1). If the indicator indicates a particular transformation, the index of the particular transformation used is signaled to the decoder on a fixed length code (2 for 4x4 blocks and 4 for 8x8 blocks in order to signal respectively 4 and 16 possible specific transformations ).

Thanks to the increase in the number of transforms, the coding performance is improved, since each transform is adapted to the statistics of a given type of residual signal.

Gains on text and graphic signals are reported by the "transform skip" technique and on signals of any kind (images containing text, synthetic graphics, or representing natural filmed scenes) by the technique presented in publication A Arrufat.

In E8, the encoder evaluates the applied transform, under control of a distortion rate criterion, from the encoded residue.

The preceding steps E1 to E8 applied to the current block c are repeated for the transformed Ts available at the encoder.

In E9, once all the available transforms have been applied, they are put into competition according to a flow-distortion criterion The quantified transformed residues associated with each candidate transform are therefore decoded to evaluate a performance measure and the transform which maximizes this criterion is finally retained.

In E10, the coded data relating to the current block are inserted in the bit stream TB.

The other blocks of image II are treated in the same way, the same for the following images of the sequence.

3. Disadvantages of the prior art

The disadvantages of the prior art are as follows:

• the multiplication of the number of transforms induces an increased signaling to indicate the transform chosen by the encoder to the decoder.

• this increased cost in signaling, which is taken into account during competition between transforms, has an impact on compression performance.

4. Objectives of the invention

The invention improves the situation.

The invention particularly aims to overcome these drawbacks of the prior art.

More specifically, an objective of the invention is to propose a solution which, by reducing the need for computing resources and the volume of data transmitted, while preserving the compression performance of a digital image coder.

5.

Statement of the invention

These objectives, as well as others which will appear subsequently, are achieved using a method of coding an image of a sequence of images, said image being divided into a plurality of blocks of processed pixels. in a defined order, said method comprising the following steps, implemented for a block, called the current block, of the current image:

Prediction of the values of the current block from at least one block previously processed from a previous or next image, called a reference image,

Calculation of a residual block by subtracting the predicted values from the original values of the current block,

Transformation of the residual block by application of a transform;

Coding of the transformed residue block;

According to the invention, the method further comprises a step of obtaining an indicator of variability of the block with respect to the at least one reference image, a step of obtaining a list of transforms among a plurality of lists predetermined at least as a function of the indicator obtained and the transformation step applies a transform from the list obtained.

The invention proposes to choose the transform to be applied to the current block in a subset of transforms adapted to the current block. The inventors have indeed found that there is a link between the variability of the current block with respect to its reference (s), and therefore between the quality of the prediction of the block and the transforms best suited to compress the information of the residue obtained. .

The invention is therefore based on an entirely new and inventive approach to image coding which consists in grouping together the best candidate transforms for coding the residue of a block as a function of information representative of a variability of the content of the block to be coded and to force the coder to make its choice in this subset.

By choosing the transform from a subset of the initial set of transforms available at the coder level, the cost of the information to be transmitted is necessarily reduced.

Advantageously, the configuration of the lists can be fixed or evolve over time.

If it is frozen, its knowledge is shared by the coder and the decoder and does not require signaling. If it is adaptive, its evolution over time must be reported. For example, the configuration of the lists can be intended for a group of images and its evolution is indicated at the level of the syntax information relating to this group or else the configuration is valid for the entire sequence of images and its evolution is indicated at the level syntax information relating to this sequence.

According to one aspect of the invention, the transformation step is repeated for the transforms of the list obtained, the method comprises a step of selecting a transform from the transforms of the list according to a distortion bit rate criterion and a coding step information identifying the transform selected in the list.

For example, the coder signals a position index of the transform selected in the list, which presupposes that the list is ordered in a similar way to the coder and to the decoder.

In a context of transform competition, having fewer transforms to test reduces the complexity of the encoder and the decoder. This advantage of the invention is added to that of having less information to report because the number of candidate transforms is reduced.

Correspondingly, the invention also relates to a method of decoding an image of a sequence of images from coded data representative of said image, said image being divided into a plurality of blocks processed in a defined order, said process comprising the following steps, implemented for a block, known as the current block:

Prediction of the current block from at least one previously processed block from at least one preceding or following image, called the reference image;

Decoding of the coded data representative of the coefficients of the transformed residual block; Reverse transformation of the transformed residue block;

Reconstruction of the decoded block from the residual block and the prediction of the current block;

According to the invention, the method comprises a step of obtaining an indicator of variability of the current block with respect to said at least one reference image, a step of obtaining a list of transforms among a plurality of predetermined lists at least as a function of the indicator obtained and a step of identifying the transform to be applied in the list obtained.

An advantage of the invention is that the steps of obtaining a variability indicator of the current block and of obtaining a list of transforms as a function of this indicator can be carried out at the level of the decoder in a manner corresponding to the coder, which reduces signaling.

According to one aspect of the invention, the number of transforms in the list obtained is a function of the block variability indicator.

With a low variability indicator, the block residue contains little energy and will be easy to compress. A small number of transforms will be sufficient. On the contrary, with a high variability indicator, one can expect a block residue comprising a lot of energy and difficult to compress. Consequently, we will choose to test a larger number of transforms. The invention thus proposes to adapt the size of the subset of transforms available for the current block to its variability, which makes it possible to optimize the signaling of the transform as a function of the current block. In a context of transform competition, the calculation costs are also reduced.

According to another aspect of the invention, the block variability indicator is a function of information representative of a time distance between the current block and the at least one reference image.

Indeed, the variability of the residuals generated by the prediction increases with the temporal distance. The subset associated with a value or a range of values of temporal distance between current image and reference image can be defined during a prior learning phase during which different groupings of transforms are tested.

For example, a list is determined comprising all the less transforms the shorter the time distance between the current image and the reference image. Indeed, the variability of the residues between the two images is supposed to be low enough for a small number of well-targeted transforms to allow the information they contain to be well compressed.

According to another aspect of the invention, the images of a group of images of the sequence being distributed between several temporal layers, the information representative of a temporal distance is evaluated on the basis of a temporal layer identifier of the image of the current block in the group.

We know the hierarchical model of temporal layers described by Wien, according to which we group in the same temporal layer the images which are at the same level in the hierarchy of images serving as references. For example, a first layer comprises the I or Intra images, which are distant from each other, but are not predicted, a second layer comprises the images predicted from the I images of the previous layer and which are inserted between the images I, a third layer comprising images which overlap in time between images of the second and images of the first layer and which are predicted with respect to images of the previous layers, etc.

It is understood that the more one goes up in the layers the more the temporal distance between the current image and its reference decreases.

Layer number information is information that is known to the encoder, as well as the decoder, and that is readily available. For example, in H.265 / HEVC, the time layer number information is transmitted in the header of each NAL unit packet (for Network Abstraction Layer), and is therefore directly accessible during decoding (see section 7.4 .2.2 of standard ITU-T H.265, "NAL unit header semantics"). The evaluation of temporal distance on the basis of such information is therefore simple and inexpensive.

According to yet another aspect of the invention, the information representative of a time distance is evaluated by difference between information representative of a time instant associated with the image of the current block and information representative of an associated time instant at least one image of the reference block.

An advantage of this information is that it is known to the coder and the decoder during the processing of a current block.

For example, the information representative of a time instant is a sequence number of the current image in the sequence. One advantage of this information is that it is easy to obtain. The difference between two image numbers or POC delta is even directly accessible, without requiring a calculation operation. For example in H.265 / HEVC the POC delta is an essential parameter during the construction of the RPS (for Referenced Picture Set) as described in section 8.3.2 of the ITU-T H.265 standard entitled HEVC, “Decoding process for referenced picture set”), and this is therefore a parameter directly accessible during encoding and decoding.

If the current block belongs to a group of blocks of an image, of slice type, its reference image (s) are defined in two lists at the level of the slice segment header (see 7.3.6.2 of standard ITU-T H.265 entitled HEVC, Referenced picture list modification syntax). One can imagine an image cut into several slices each having separate lists. It is therefore possible that two slices of the same image do not use the same reference images for their blocks, and therefore that the time distance information does not take the same value.

According to another aspect of the invention, the prediction step comprises obtaining a motion vector between the current block and a block of the reference image and the variability indicator of the block is obtained from at least one characteristic of the motion vector obtained.

The amplitude or the precision of quantification of the motion vector are information from which one can evaluate the variability of the content to be coded compared to its reference. The amplitude of the motion vector indicates whether the content to be coded in the current block has moved a lot compared to the reference image. Likewise, the level of quantization precision is chosen by the coder during the interpolation of the reference image (for example 1/4, 1/8 or even 1/16 pixels).

According to another aspect of the invention, the list comprising several transforms, the decoding method further comprises a step of decoding information identifying the transform in the list.

For example, when the list includes an ordered sequence of transforms, the identification information is an index of position of the transform in the sequence.

The various embodiments or features mentioned below can be added independently or in combination with each other, to the features of the decoding method and / or of the encoding method defined above.

The invention also relates to a coding device suitable for implementing the coding method according to any one of the particular embodiments defined above. This coding device could of course include the various characteristics relating to the coding method according to the invention. Thus, the characteristics and advantages of this coding device are the same as those of the decoding method, and are not described in more detail.

The invention also relates to a decoding device suitable for implementing the decoding method according to any one of the particular embodiments defined above. This decoding device could of course include the various characteristics relating to the decoding method according to the invention. Thus, the characteristics and advantages of this decoding device are the same as those of the decoding method, and are not described in more detail.

Correlatively, the invention relates to a signal carrying a bit stream comprising coded data representative of an image of a sequence of images, said image being divided into blocks of pixels processed in a defined order, the values of a current block being predicted from the values of at least one block previously processed from at least one reference image, the values of a residual block being calculated by subtracting the predicted values from the original values of the current block, a transformed residual block being obtained by applying a transform to pixels of the residual block;

According to the invention, said signal comprises coded information representative of a position index of the transform to be applied to the current residual block in a list comprising an ordered sequence of transforms, the list being intended to be obtained at least as a function of a variability indicator of the current block with respect to said at least one reference image.

Correlatively, the invention also relates to terminal equipment comprising a transmission module and a data reception module to and from a telecommunications network, a coding device and a decoding device according to the invention.

The invention also relates to a computer program comprising instructions for implementing the steps of a method for coding a digital image as described above, when this program is executed by a processor.

The invention also relates to a computer program comprising instructions for implementing the steps of a method for decoding a digital image as described above, when this program is executed by a processor.

These programs can use any programming language. They can be downloaded from a communication network and / or saved on a computer-readable medium.

The invention finally relates to recording media, readable by a processor, integrated or not in the device for coding a digital image and in the device for decoding a digital image according to the invention, optionally removable, storing respectively a computer program implementing a coding method and a computer program implementing a decoding method, as described above.

6. List of figures

Other advantages and characteristics of the invention will appear more clearly on reading the following description of a particular embodiment of the invention, given by way of simple illustrative and nonlimiting example, and the appended drawings, among which :

Figure 1 (already described) schematically shows a sequence of digital images cut into blocks of pixels;

FIGS. 2A and 2B (already described) show the DCT and DST transforms implemented by the encoder of the HEVC standard;

FIG. 3 schematically presents the steps of a method for coding a digital image according to the invention;

FIG. 4 schematically presents the steps of a method for decoding a digital image according to the invention;

FIG. 5 schematically presents a model of hierarchical temporal structure of a group of images implemented by an embodiment of the invention;

FIG. 6 schematically shows a motion compensation implemented during the prediction in inter mode of a current block with respect to one or more past or future reference images according to the prior art;

FIG. 7 presents a first example of a transform table associating a time layer number with a list according to an embodiment of the invention;

FIGS. 8A and 8B present a second example based on two tables associating for the first table a time layer number with a number of transforms, to be identified in a list of transforms indicated in the second table, according to a second embodiment of the invention;

FIG. 9 presents examples of configurations of lists of transforms as a function of a time layer number;

FIG. 10 shows the performances achieved by an encoder implementing the encoding method according to the invention, when the variability indicator used is a time layer number;

FIG. 11 schematically presents the hardware structure of a device for coding a digital image according to the invention; and FIG. 12 schematically presents the hardware structure of a device for decoding a digital image according to the invention.

7. Description of a particular embodiment of the invention

The general principle of the invention is based on a prior association of sub-sets of transforms with values or ranges of values of a variability indicator of the current block with respect to at least one reference image. This indicator is estimated similarly to the coder and the decoder. The encoder obtains an indicator value for the current block, chooses the transform to be applied to the current block in the subset of transforms associated with this value of the indicator. If necessary, it then signals to the decoder which transform it has used in this subset. Correspondingly, the decoder obtains a variability indicator value for the current block and accesses the transform subset associated with it. Using the possible signaling, it identifies the transform applied to the current block and applies the reverse transform.

In the following description, we consider an original video made up of a sequence of K images Ii, I2, ... Ik, with K non-zero integer, such as that already presented in relation to Figure 1. The images are encoded by an encoder, the encoded data is inserted a binary train TB transmitted to a decoder via a communication network, or a compressed file FC, intended to be stored on a hard disk for example. The decoder extracts the coded data, then received and decoded by a decoder in a predefined order known to the encoder and the decoder, for example in the time order Ii, then I2, ..., then IK, this order may differ according to the embodiment.

In the following description, we place ourselves in a context of transform competition, according to which the coder tests the transforms of a subset of transforms and chooses for the current block the one which achieves the best compression performance according to a criterion. flow-distortion.

7.1 Method for coding an image of a sequence

In relation to Figure 3, we now consider the steps of a coding method according to an embodiment of the invention. In the following description, we place ourselves in particular in a context of transform competition.

An Ik image is cut into CTU blocks (for “Coding Tree Unit”, in English) of size, for example equal to 64 × 64 pixels.

During a step E0, a block to be processed is selected, called the current block c. For example, it is a CU block (for “Coding Unit”, in English), square or rectangular, of dimensions MxN, with M and N integers not harmful, obtained by partitioning of a CTU block.

During a step E1, a step of predicting the block c is carried out. This operation, in accordance with the prior art, is carried out from pixels originating from the image being coded (intra coding) or based on an image already processed by the encoding operation (inter coding). We get a predicted block Pr.

In the following description, it is assumed that the current block is predicted according to an inter prediction mode, that is to say with respect to one or more previously processed images, called reference images.

During a step E2, the current block of pixels is subtracted, pixel by pixel, from the block predicted during the previous operation. We obtain a block of residual pixels R.

According to the invention, an indicator of variability IV of a prediction of the current block with respect to the reference image (s) is obtained in E3. Several embodiments of this step will be detailed below.

During a step E4, the possible transforms are identified for the prediction mode considered and the size of the current block. The term “possible transforms” denotes a plurality T of transforms available in the memory of the encoder in a preset manner, with T non-zero integer, for example T is equal to 32. For example, this plurality is stored in the form of an ordered list L or not, transforms. Note that this list can be predetermined or adaptive.

According to the invention, a plurality of lists L1 to LJ, with nonzero integer J, has been formed from the available transforms and one of these lists is associated with the value of variability indicator IV obtained. For example, a table stored in memory includes records EGj, with j an integer less than T which each correspond to a particular value or range of values of this variability indicator, a list Lj coming from the list L. Advantageously, a EGj record of the table associates with each value or range of values of the indicator, a pointer to this list Lj. Examples of tables will be presented below in relation to Figures 7 and 8.

In some cases, the configuration of the lists associated with the different values of variability indicators may vary over time, to adapt to the change in content of the sequence. For example, the table includes a first record comprising a first list associated with a predetermined value or range of values of the variability indicator. It is assumed that the coder decides, following a change of scene, to replace the first record by a second record comprising a second list, associated with the same range of values. This update must be signaled in the bit stream so that the decoder updates its own table.

During a step E5, a particular Tri transform is identified in the list Lj obtained.

During a step E6, for the identified transform Τη, with i non-zero integer between 0 and T-1, the residual signal R is transformed. A residual residual signal RT, is obtained.

In E7, as is known in the state of the art, the coefficients of the transformed residual block are scanned in a predetermined order so as to constitute a one-dimensional vector RTi [m], where the index m varies from 0 to Nb- 1, with Nb number of pixels in the current block c. The index m is called the frequency of the coefficient R [m]. Conventionally, these coefficients are scanned in globally increasing or decreasing order of frequency values, for example according to a zigzag path, which is known from the image coding standard. fixed JPEG. This route mode (for “scanning” in English) can also depend on the transform applied. As the path mode necessarily influences the final order of the coefficients transformed in the vector RTi [m], we consider below that a transform identified Τη in the list Lj is associated with a particular path mode. In other words, the same transform associated with another path mode will be assigned another transform identifier in the list Lj and will therefore be considered as a distinct transform.

The components of the vector RTi are then quantified according to a given quantification method, scalar or vector known to those skilled in the art, with a quantization parameter QP regulating the precision of the approximation carried out in this step. An RTQi quantified vector is obtained.

In E8, the information relating to the current block, comprising in particular the quantified data and elements of description of the block such as the prediction mode, are encoded by a known entropy coding technique, such as for example Huffman coding, arithmetic coding or CABAC coding as used in the HEVC standard.

During this step, it is examined whether it is necessary to report identification information of the Tri transform applied to the current block. There are two cases:

When the list Lj includes only one transform identifier, it is not necessary to report anything. Indeed, as we will see below, the decoder according to the invention is arranged to obtain the list Lj correspondingly to that of the encoder. An advantage of this embodiment is that it is simple to implement and that it does not require signaling, insofar as the configuration of the list is fixed.

When the list Lj comprises more than one transform identifier, the coder signals to the decoder a position index POS of the transform Tri applied to the current block in the list Lj. To do this, the list is advantageously ordered so that the coder and the decoder share the knowledge of the positions of the transforms in the list. For example, an identifier of the Tri transform in a list of four elements is coded on 4 bits (transforms TrO, tri, Tr2 and Tr3 respectively for a code 00, 01, 10, 11) while a list of 8 elements would cost 3 bits (000, 001, 010, 011, 100, 101, 110 and 111 respectively for the transforms TrO, Tri, Tr2, Tr3, Tr4, Tr5, Tr6, Tr7 and Tr8), in the case of a fixed length code .

An advantage of the invention is that few bits are necessary to signal the Tri transforms in the binary train, contrary to the state of the art which explicitly transmits its complete identifier.

In E9, the coding performance of the current block, for this Tri transform, is evaluated, in a manner known to those skilled in the art, from a cost function. The transform chosen is that which maximizes the compression performance, according to a distortion rate criterion.

For example, through a Lagrangian cost measure J = D + À x R, where D is the distortion measured on the reconstructed block, R the bit rate generated during coding and À the Lagrange multiplier.

In E10, it is checked whether all the transforms of the list Lj have been applied.

If this is not the case, we return to step E5 during which another transform of the list Lj is identified. Steps E6 to E10 are repeated for the new transform.

If this is the case, the best transform is decided during step Eli, on the basis of the cost function values obtained by each of the transforms of the list Lj and according to a flow-distortion optimization criterion.

It will be noted that in the case of a CTU block, the competition of transforms is nested with the selection of the best division into sub-blocks CU, that is to say that during the step Eli the best is chosen. transformed for each of the sub-blocks CU of the current block CTU.

The coded data corresponding to the chosen transform are inserted into the bit stream or file compressed in E12.

In E13, we test if there are still blocks to process in the current image. If so, the process continues with the step EO of selecting a block C to be processed. Otherwise, the processing is finished for this image and we go to the next image according to the encoding order.

The compressed file or bit stream produced by the coding method which has just been described is transmitted to a decoder for example via a telecommunications network.

7.2 Image decoding process of a sequence of images

It is assumed that the bit stream TB has been received by a decoding device implementing a decoding method according to the invention. This decoding process will now be described in relation to Figure 4.

In DO, we start by selecting as current block C 'the first block to be processed. For example, this is the first block (in lexicographic order). This block contains MxN pixels, with M and N integers not harmful.

As described for the encoding method, the block C ′ considered can be a CTU block or a CU sub-block obtained by cutting out the CTU block or a residue block or sub-block obtained by subtracting a prediction from the current block to the current block.

During a step D1, the coded data relating to the current block C 'are read and decoded. The coded data includes coding parameters, such as for example the prediction mode used and the values relating to the amplitudes and to the signs of the quantized residual coefficients of the current block.

When the prediction mode determined indicates that a prediction has been made by the encoder, the current block is predicted in D2, according to the prediction mode determined from a block already processed. A predicted block Pr 'is obtained.

During a step D3, the read data representative of the quantized, residual values of the current block (values and signs of the coefficients) is decoded, in the form of a vector of values RQ ′. It is understood that this is the opposite operation to that of entropy coding previously described in relation to the encoding method.

During a step D4, the coefficients of the vectors RQ ′ are dequantized using a quantization step predetermined or read from the compressed file.

In D5, a transformed block RT ′ is reconstructed from the dequantified vector.

In D6, according to the invention, an indicator of variability IV of the current block is obtained with respect to its reference (s), corresponding to that of the coder. Several embodiments will be detailed below. It will be seen that, depending on the implementations considered, the obtaining can be carried out once for all the blocks of the same image or of the same group of images, or else must be repeated for each current block.

In D7, we recover a list of transforms Lj associated with the value of the indicator obtained. For example, when the configurations of the lists are predetermined, they are known to the decoder and associated respectively with a value or a range of values of the indicator. If the configurations are not predetermined, they are signaled in the bit stream prior to the decoding of the image so as to be known to the decoder.

In D8, we identify the Tri transform applied to the residual block by the encoder. Two cases are considered:

the list Lj contains only one transform. The decoder accesses it directly without the need for additional information;

the Lj list contains more than one transform. The decoder reads signaling information from the Sort transform from the list in the binary train or the compressed file. This is for example a POS position index of the transform in the list. The underlying hypothesis is that the coder and the decoder share the same knowledge of an order of the transforms in the list. .

Correspondingly to the coding, it is noted that the decoder may have to identify a Sort transform by sub-block CU of the current CTU block.

In D9, the inverse transform of Tri is applied to the residue block R '. A residual block of pixels r 'is obtained.

In D10, the block c 'is reconstructed from its prediction Pr' and the residue block r '. It is stored in a memory M1, so that it can serve as a prediction for a next block.

During a step D12, we come to test if the current block is the last block to process the decoder, taking into account the order of traversal defined previously. If so, the decoding process has finished processing the current image and goes to the next image. If not, the next step is the step of selecting the next block DO and the decoding steps D1 to D12 previously described are repeated for the next block selected.

7.3 Obtaining a variability indicator for a current block

We will now describe several embodiments of step E3 of obtaining a variability indicator of a current block C and of step E4 of obtaining a list as a function of this indicator. Correspondingly, these embodiments apply to steps D6 and D7 of the decoding method according to the invention.

7.3.1 Obtaining information representative of a time distance between the current block and at least one reference block

According to a first embodiment of the invention, the variability indicator is obtained from information representative of a time distance between the current block and its reference image or images.

In relation to Figure 4, we consider a hierarchical representation model in time layers, known as the Wien model, of a group of images or GoP (for “Group of Pictures”, in English) of the sequence, typically set works by successive video compression standards, such as for example H265 or HEVC.

In the example shown, the GoP includes 9 images. In the input sequence, the GoP images are ordered from 10 to 18 with associated time instants such as tOctl <... <t8, corresponding to an order of restitution of the video images. When encoding, these images are processed according to an encoding order which differs from the restitution order. The temporal encoding structure is based on different types of images, among which there are:

Intra images, which are coded independently of other previously processed images and serve as a reference for the inter prediction of other GoP images; Inter type P images, which are coded by temporal prediction from a reference image already processed, which may correspond to a past or future restitution instant; an image P can itself serve as a temporal reference to another image; Inter type B images, which are coded by time prediction from two reference images, a past and a future. An image B can serve as a time reference to another image.

In the example shown, 10 and 18 are Intra images. Images II to 17 are images B. There is no image P.

According to the Wien model, the GoP is broken down into several time layers:

a first layer tidO, comprising the images 110 and 18. Image 10 is coded first, so it has the coding rank rcO = 0 and image 18 is coded second, its rank is rc8 = 1;

a second layer tidl, comprising the image 14 of type B. Its reference images are 10 and 18. Its coding rank is rc4 = 2;

a third layer tid2, comprising images 12 and 16 of type B. 12 has references 10 and 14. 16 has references 14 and 18. Their coding ranks are rc2 = 3 and rc6 = 4; a fourth layer tid3 comprising images II, 13, 15 and 17. It has references 10 and 12, its coding rank is rcl = 5; 13 has references 12 and 14, its coding rank is rc3 = 6; 15 has references 14 and 16, its coding rank is rc5 = 7; 17 has references 16 and 18, its coding rank is rc 7.

It can be seen that an image is all the more distant from its references that it belongs to a lower temporal layer.

The inventors have found that the more temporally distant the current image from its references, the more the variability of the block increases. Indeed, the probability that the content of the current image has changed compared to that of its references is then all the greater.

According to this embodiment of the invention, the block variability indicator comprises the time layer number to which the current image belongs. An advantage is that this value is available at the encoder level and very easy to obtain on the decoder side because, according to the ITU-T H265 / HEVC standard for example, this information is specified in the syntax element nuh_temporaljd_plusl located in the nal_unit_header () header as specified in section

7.3.1.2 of the ITU-T H.265 specification entitled "High Efficiency Video Coding", published in April 2015.

The time layer number is common to all the blocks of the same image or of a subset of this image, for example of the “slice” type. However, this value is read from a memory for each block to be processed. On the encoder side, step E3 of obtaining a variability indicator is therefore repeated for each current block. On the decoder side, even if the time layer number value is read from the bitstream once and for all, it is stored in memory, so that the decoder can access it for processing the current block.

In E4, we therefore obtain a list Lj of transforms associated with the time layer number tidj of the current block.

We therefore previously created a specific transform list for each time layer number. For example, a lookup table maps a given time level to such a list. This correspondence is either fixed (known to the coder and the decoder), or adaptive. The adaptive case involves signaling the necessary information, for example a pointer to a new list. This information will for example be placed and transmitted for a given sequence of images, for example in the syntax structure SPS (for Sequence Parameter Set, in English) described in section 7.3.2.2 of the specification ITU-T H.265, or also in the PPS syntactic structure for (picture parameter set, in English), described in section 7.3.2.3 of the same standard. These syntactic structures are transmitted to the decoder in the binary train prior to the decoding information of the image sequence ut (comprising the coded data representative of the residues, motion vectors, prediction modes, filters, etc.).

The step of obtaining the list Lj as a function of the time layer number is therefore also carried out once for all the blocks of the image or of the slice considered.

We now detail an example of implementation of this embodiment of the invention in the context of JEM-3.0, which is the test software currently used in the JVET standardization group (for Joint Video Exploration Team, joint group formed by experts from MPEG and VCEG, respectively attached to ISO / IEC and ITU). This software is described in the document JVET-C1001 at the address http://phenix.it-sudparis.eu/ivet/doc end user / documents / 3 Geneva / wqll / JVETC1001-v3.zip.

Consider as examples 4 distinct transform lists, of increasing sizes:

. LO: DCT-2. Ll: DCT-2 + {TO, Tl}. L2: DCT-2 + {TO, Tl, T2, T3}. L3: DCT-2 + {TO, Tl, T2, T3, T4, T5, T6, T7}.

In relation to Figure 7, an example of a list association with a time layer number is presented: tid3 is associated with LO, tid2 with L1, tidl with L2 and tidO with L3.

Alternatively, these lists can be obtained from the tables of Figures 8A and 8B. Figure 8A associates with a list identifier a number of transforms, for example 1 for L3, 3 for L2, 5 for L1 and 7 for LO and Figure 8B comprises a table composed of several rows, a row or record of the table associating with particular values of block size an ordered sequence of transformed T1s. This table is advantageously stored in memory of the coder or of the decoder.

We see that the lists Lj are obtained from this single table.

The transformed T-1s are arranged in an order inversely proportional to their frequency of belonging to the preconfigured lists. In other words, the first two transforms of the sequence are part of all 4 lists, the next two from the last 3 lists L2, Ll, LO, the next two from the last 2 lists L1, LO and the last two from the largest LO list. A first transform, in this case the DCT2-2D is common to all the lists and is not indicated in the table.

It is understood that to recover the identifiers of the transforms of a list Lj, it suffices in this embodiment to go and read in the table of FIG. 8A the number of transforms which it comprises and to read the number of corresponding identifiers in the correct row of the table of figure 8B

Note that the table in Figure 8B indicates pairs of transforms rather than single transforms, which corresponds to the fact that in image and video coding, the most used transforms are generally linear and orthogonal block transforms, also applicable. both in rows and columns of the current block. In this case one specifies a couple of transforms L-C which apply respectively to the rows and the columns of the residue according to the relation expressed in the equation below:

X - L - (c -xf, with x residual pixel block and X transformed residual block.

In relation to the table in Figure 9, we now consider examples of list configurations by time layer. These different configurations are made available to a JEM-3 type encoder:

These different configurations have been tested in the JEM-3.0 coder and decoder and have led to the results of Figure 10. It can be seen that a suitable list configuration makes it possible to improve the coding performance of the JEM-3.0 coder. Here, the configuration is updated for each new sequence.

For example, for the ClassD-S02-BQSquare_416x240_60 sequence, the following configurations are used:

• for quantization steps QP22 and QP32, use of configuration 0;

• for the quantification step QP27 -a use of configuration 2;

• for the quantification step QP37 -a use of configuration 5.

On the sequences tested, this embodiment of the invention makes it possible to improve on average bit rate by 0.14% with higher results on certain sequences (up to -0.33%). In this example, the configuration of the lists and their association to the time layer numbers by sequence is signaled in the SPS header.

It is noted that an increase in the level of granularity (for example by passing to the image or PPS level) with a more regular updating of the lists would make the system even more efficient, since it is more adapted to temporal variations in content and therefore more to deal with the variety and type of residues encountered. For example, if the encoder evaluates that the list associated with a time layer is no longer suited to the content it encodes, for example because it has detected a change of scene, then it can choose to signal on the train binary, for example at the level of the PPS header, the information necessary for updating the lists, so that the decoder is configured to decode and interpret.

Advantageously, the coder systematically inserts a bit into the PPS to indicate to the decoder whether an update of the lists associated with the time layer numbers should be made or not. If this bit is at 1, then the following signaling must be decoded (example for the case where J = 2 ^N lists are available):

• tidO time layer • tidl time layer • tid2 time layer • tid3 time layer

N bits to signal the list number. N bits to signal the list number. N bits to signal the list number. N bits to signal the list number.

According to a second embodiment of the invention, the information representative of a time distance between the current block and the reference block corresponds to a number of images separating the current image from its reference.

This information is directly known to the coder and it is signaled to the decoder under the parameter name Delta POC for difference in rank between images (Delta Picture Order Count, in English) in a header of the compressed file or of the bit stream. It is notably transmitted for a given sequence of images, on the one hand in the sequence parameter set (SPS, see

7.3.2.2 of the ITU-T H265 standard), or on the other hand in the slice_segment_header (see 7.3.6.1 of the ITU-T H265 standard). More precisely, this information is derived directly during the decoding of the short-term referenced picture set (st_ref_pic_set, see 7.3.7 of the ITU-T H265 standard), which is included in the SPS and the slice_segment_header. An advantage of this information is that it is directly available, without requiring prior calculation.

As a variant, it is possible to calculate a time distance by difference between the time instants associated respectively with the current image and with the reference image.

For an image B, two references are used. We can for example make an average between the two temporal distances obtained or even keep the largest distance.

This measurement of temporal distance is common to all the blocks of the same image and, where appropriate, of the same slice, of which all the blocks share the same references. It can therefore be obtained once for all the blocks of the same image / slice and stored in a memory to which the coder accesses for processing a current block.

In this second embodiment of the invention, lists are therefore associated with the different values of temporal distances obtained.

This correspondence is made beforehand, from a learning based on a representative set of sequences. It can then be fixed and in this case known to the coder and the decoder or else be adaptive, which requires appropriate signaling. The encoder can choose to transmit the information necessary for updating the correspondence table in the bit stream.

The coder obtains the list associated with the measurement of the temporal distance of the image / slice with its reference or references and uses it for all the blocks which it comprises.

For each block of the image, the coder successively applies the transforms of the list obtained to its residue, chooses the best transform and codes an index corresponding to a position of the transform chosen in the list. :

7.3.2 Obtaining a predictive variability indicator as a function of a characteristic of a motion vector

In the case of an Inter prediction, the prediction step E 1 comprises, in a known manner, a substep for estimating a motion vector from the current block to an area of the reference image, of the same size as the block, according to a similarity criterion and a sub-step of compensation for the movement of the current block with respect to the reference image as a function of the estimated vector.

In FIG. 6, a first motion vector MV1 is estimated, estimated between the block Bl of a reference image IR1, collocated with the current block C of the current image Ik and a zone ZI of the image IR1. This motion vector is then used to compensate for the movement of the collocated block B1 with respect to the current block C. In a similar manner, a second motion vector MV2 has been estimated estimated in the current block B2 of a reference image IR2 collocated with the block current C and an area Z2 of the image IR2. MV2 is used to compensate for the movement of block B2 with respect to C.

In the bidirectional case, an indicator of global variability is estimated. The indicators are estimated individually for each motion vector and one is chosen. It is possible, for example, to choose the lowest indicator value as the global indicator value, for example the motion vector of lower amplitude or that which requires the finest quantization precision. The calculated motion vector is then quantified and then coded in the compressed file or binary train.

As a variant, an indicator of variability can be constructed from the estimated motion vectors, for example as the average of these vectors.

The invention proposes to use the knowledge of characteristics of this motion vector to estimate an indicator of variability for predicting the current block.

It is noted that the decoder accesses the motion vector used for the prediction of the current block, since it is transmitted in the binary train. It can therefore perform the same operations as the encoder.

According to a third alternative embodiment, the prediction variability indicator for the current block is obtained from information representative of an accuracy of the motion vector used during inter prediction. This is for example the interpolation precision required by the displacement induced by the motion vector (for example ¹ A pixel, 1/8, 1/16, etc.).

According to a fourth embodiment of the invention, the variability indicator is obtained from the amplitude of the displacement induced by the motion vector during the prediction. In this case, this means that the content of the image has moved a lot between the reference image and the current image and that the prediction variability is high. For example, this amplitude is compared with one or more thresholds, depending on the number of lists created. The higher the amplitude of the motion vector, the higher the variability indicator and the more difficult it is to predict. Advantageously, the list of transforms associated with a range of high amplitude values comprises a number of transforms greater than that of a list associated with low amplitude values.

The motion estimation being performed for each current block in inter mode, step E3 of obtaining the interpolation precision information of the motion vector must be repeated for each block in inter mode. It will be noted that according to the HEVC standard, all the blocks of an inter slice are not necessarily in inter mode. Certain blocks can be coded in intra mode and in this case the invention does not apply to them.

For example, lists of transforms are associated beforehand with the different possible values of precision for the motion vectors, such as for example:

• 1/4 pel * · L3: DCT-2 + {TO, Tl, T2, T3, T4, T5, T6, T7}.

• 1/8 pel * · L2: DCT-2 + {TO, Tl, T2, T3} • 1/16 pel * · Ll: DCT-2 + {TO, Tl} • 1/32 pel * LO: DCT- 2

The higher the precision of the motion vectors, the more we can expect the prediction to be correct and therefore the less it is necessary to test transforms.

If the configuration of the lists is fixed, it is known to the coder and the decoder. If it varies over time, then a new configuration is transmitted in the signal for example, for a given image sequence, in the SPS part (see 7.3.2.2 of the standard ITU-T H. 265) or in PPS (see 7.3.2.3 of standard ITU-T H. 265). The decoder therefore obtains the p

For example, if the coder has at its disposal 2 ^N sublists, with N integer, the following information is successively coded:

• Precision 1/4 -> N bits to signal the list number.

• Accuracy 1/8 -> N bits to signal the list number.

• Accuracy 1/16 -> N bits to signal the list number.

• Accuracy 1/32 -> N bits to signal the list number.

In the case where the coder would have available N = 4 lists, with the allocation 1/4 - »L3, 1/8 -» L2, 1/16 -> Ll and 1/32 -> LO, we would indicate the following information in the SPS: 0000-1111-01001100.

Of course, the invention is not limited to the embodiments which have just been described. They can in particular be combined with each other, for example to obtain an indicator of variability for predicting the current block both from information representative of a time distance between the current block and the reference image and of a characteristic information of the motion vector estimated for the block. For example, the variability indicator is constructed as a weighted sum of the two values or else it includes this pair of values to which a list of transforms is associated. An advantage is to combine information common to all the blocks of an image / slice (for example the number of time layer) with information specific to the current block (for example a characteristic of its motion vector) and thus render the value of the even more relevant variability indicator.

The coding and decoding methods described above can be integrated into standard video coders / decoders such as HEVC / H.265, AVC / H.264, coders / decoders of a future standard or any type of coders / decoders video owners. The coding and decoding methods according to the invention also apply to all types of signals using predictive coding.

FIG. 11 shows the simplified structure of a coding device 100 adapted to implement the coding method according to any one of the particular embodiments of the invention, which have just been described in relation to FIGS. 3 and 5. The coding device 100 is suitable for coding at least one image cut into blocks in the form of a bit stream TB or compressed file FC.

The coding device 100 is notably configured to:

predict values of the current block from at least one previously processed block of a previous or next image, called the reference image, calculate a residual block (R) by subtracting the predicted values from the original values of the current block, transforming a block residue by application of a transform, the transform belonging to a predetermined list of transforms;

coding the transformed residue block and of information identifying the applied transform;

According to the invention, the device is further configured to:

obtaining a predictive variability indicator of the block with respect to the at least one reference block;

obtaining a list of transforms from among a plurality of predetermined lists at least as a function of the indicator obtained; and apply an identified transform belonging to the list obtained.

According to a particular embodiment of the invention, the steps of the coding method are implemented by computer program instructions. For this, the coding device 100 has the conventional architecture of a computer and in particular comprises a memory MEM1, a processing unit UT1, equipped for example with a microprocessor pl, and controlled by the computer program Pgl stored in MEM1 memory. The computer program Pgl includes instructions for implementing the steps of the coding method as described above, when the program is executed by the processor μΐ.

On initialization, the code instructions of the computer program Pgl are for example loaded into a RAM memory before being executed by the processor PROC. The processor p1 of the processing unit UT1 implements in particular the steps of the coding method described above, according to the instructions of the computer program Pgl.

According to another particular embodiment of the invention, the coding method is implemented by modules or functional units (s). For this, the coding device 100 further comprises the following modules:

Prediction of the values of the current block from at least one previously processed block of a previous or next image, called the reference image,

Calculation of a residual block by subtracting the predicted values from the original values of the current block,

Transformation of the residual block by application of a transform;

Coding of the transformed residue block;

According to the invention, the device further comprises a unit for obtaining a predictive variability indicator of the current block, a unit for obtaining a list of transforms among a plurality of predetermined lists at least from l 'indicator obtained and the transformation unit applies a transform from the determined list.

The processing unit UT1 cooperates with the various functional modules described above and the memory MEM1 in order to implement the steps of the coding method.

According to one embodiment of the invention, the device 100 further comprises a unit Mi for storing the tables previously described in the various embodiments of the invention, such as for example a table comprising records associating with a value or a range of values for the variability indicator a list of transforms.

These units are controlled by the processor μι of the processing unit 110.

Advantageously, such a coding device 100 can be integrated into user terminal equipment TU, such as a personal computer, a tablet, a digital camera, a smart mobile phone (for “smartphone”, in English), etc. The device 100 is then arranged to cooperate at least with the following module of the terminal TU:

a data transmission / reception E / R module, via which the bit stream TB or the compressed file FC is transmitted in a telecommunications network, for example a wired, radio, or wireless network.

The different functional modules described above can be in hardware and / or software. In software form, such a functional module can comprise a processor, a memory and program code instructions for implementing the function corresponding to the module when the code instructions are executed by a processor. In a physical form, such a functional module can be implemented by any type of suitable encoding circuits, such as for example and without limitation microprocessors, signal processing processors (DSP for Digital Signal Processor in English) , application-specific integrated circuits (ASICs for Application Specifies Integrated Circuit in English), FPGA circuits for Field Programmable Gate Arrays in English, wiring of logic units.

FIG. 12 shows the simplified structure of a decoding device 200 adapted to implement the decoding method according to any one of the particular embodiments of the invention which have just been described in relation to FIGS. 4 and 5. The decoding device 200 is suitable for decoding a bit stream or a file comprising coded data representative of at least one image, said image being divided into blocks. The decoding device 200 is notably configured to:

Predict a current block from at least one previously processed block from a previous or next image, called the reference image;

Decode coded values of the coefficients of the transformed residual block extracted from the binary train;

Carry out a reverse transformation of the transformed residue block;

Rebuild the decoded block from the residual block and the prediction of the current block;

According to the invention, it is further configured to obtain an indicator for predicting variability of the current block with respect to at least one reference image, obtaining a list of transforms from a plurality of predetermined lists at least according to the indicator. obtained and identify the transform to apply in the determined list.

According to a particular embodiment of the invention, the decoding device 200 has the conventional architecture of a computer and in particular comprises a memory MEM2, a processing unit UT2, equipped for example with a microprocessor μ2, and controlled by the computer program Pg2 stored in memory MEM2. The computer program Pg2 includes instructions for implementing the steps of the decoding method as described above, when the program is executed by the processor μ2.

At initialization, the code instructions of the computer program Pg2 are for example loaded into a RAM memory before being executed by the processor μ2. The processor μ2 of the processing unit UT2 implements in particular the steps of the decoding method described above, according to the instructions of the computer program Pg2.

According to another particular embodiment of the invention, the decoding method is implemented by functional modules. For this, the decoding device 200 further comprises the following modules:

Prediction of a current block from at least one previously processed block of a previous or next image, called the reference image;

Decoding of the coded values of the coefficients of the transformed residual block extracted from the binary train;

Inverse transformation of the transformed residual block by application of a transform;

Reconstruction of the decoded block from the residual block and the prediction of the current block;

According to the invention, it further comprises a unit for obtaining an indicator of variability for predicting the current block with respect to at least one reference image, a unit for obtaining a list of transforms among a plurality of predetermined lists at least as a function of the indicator evaluated and a unit of identification of the transform in the list obtained.

The processing unit μ2 cooperates with the various functional modules described above and the memory MEM2 in order to implement the steps of the decoding process.

The device 200 further comprises a unit M ₂ for storing a table associating with a value or range of value of the variability indicator a list of transforms.

These units are controlled by the processor μ2 of the processing unit μ2.

The different functional modules described above can be in hardware and / or software. In software form, such a functional module can comprise a processor, a memory and program code instructions for implementing the function corresponding to the module when the code instructions are executed by a processor.

In a physical form, such a functional module can be implemented by any type of suitable encoding circuits, such as for example and without limitation microprocessors, signal processing processors (DSP for Digital Signal Processor in English) , application-specific integrated circuits (ASICs for Application Specifies Integrated Circuit in English), FPGA circuits for Field Programmable Gate Arrays in English, wiring of logic units.

Advantageously, such a device 200 can be integrated into a user terminal TU, for example a decoder, a TV connection box (for “Set-Top-Box”, in English), a digital television set, a computer, tablet, smart phone, etc. The device 200 is then arranged to cooperate at least with the following module of the terminal TU:

a data transmission / reception I / O module, through which the bit stream

TB or the FC compressed file is received from the telecommunications network.

a DISP module for displaying decoded digital images.

The invention which has just been presented can be used in any image or video coding system. In particular, it is intended to be valued in a future ITU / MPEG compression standard. It can find many applications, requiring compression of video signal, audio signal (speech, sound), still images, images acquired by a medical imaging module. It applies for example both to two-dimensional (2D), three-dimensional (3D) content including a depth map, or even multispectral images (whose color intensities are different from the three red green blue bands) or finally full images. The invention.

It goes without saying that the embodiments which have been described above have been given for purely indicative and in no way limitative, and that numerous modifications can be easily made by those skilled in the art without departing from the scope. of the invention.

Claims (15)

1. A method of coding an image of a sequence of digital images, said image being divided into a plurality of blocks of pixels processed in a defined order, said method comprising the following steps, implemented for a block, known as current block, of the current image, of predetermined dimensions: Prediction (El) of the values of the current block from at least one previously processed block originating from at least one preceding or following image, called the reference image, Calculation (E2) of a residual block (R) by subtracting the predicted values from the original values of the current block, Transformation (E6) of the residual block by application of a transform; Coding (E7) of the transformed residue block; characterized in that the method further comprises a step (E3) of obtaining an indicator of variability of the block with respect to the at least one reference image, a step of obtaining (E4) of a list of transforms among a plurality of lists at least as a function of the indicator obtained and in that the transformation step applies a transform of the list obtained. 2. Method for decoding an image of a sequence of images from coded data representative of said image, said image (Ik) being divided into a plurality of blocks processed in a defined order, said method comprising the following steps , implemented for a block (C '), known as the current block: Prediction (D2) of the current block from at least one previously processed block of at least one previous or next image, called the reference image; Decoding (D3) of the coded data representative of the coefficients of the transformed residual block; Inverse transformation (D9) of the transformed residue block; Reconstruction (D10) of the decoded block from the residual block and the prediction of the current block; said method being characterized in that it comprises a step (D6) of obtaining an indicator of variability of the current block with respect to said at least one reference image, a step of obtaining (D7) of a list transforms among a plurality of predetermined lists at least as a function of the indicator obtained and an identification step (D8) of the transform to be applied in the list obtained. 3. Method according to one of claims 1 or 2, characterized in that the number of transforms from the list obtained is a function of the block variability indicator. 4. Method according to one of claims 1 to 3, characterized in that obtaining the block variability indicator comprises obtaining information representative of a time distance between the current block and said at least a reference image. 5. Method according to claim 4, characterized in that, the images of a group of images of the sequence being distributed between several time layers, the information representative of a time distance obtained from a time layer identifier of the image of the current block in the group of images. 6. Method according to claim 4, characterized in that the information representative of a time distance is obtained by difference between information representative of a time instant associated with the image of the current block and information representative of a time temporal associated with said at least one reference image. 7. Method according to one of claims 1 to 3, characterized in that, the prediction step comprising a sub-step of obtaining a motion vector between the current block and the reference image, the indicator variability of the block is obtained from at least one characteristic of the motion vector. 8. Image coding method according to one of claims 1 and 3 to 7, characterized in that, when the list obtained comprises an ordered sequence of several transforms, the transformation step is repeated for the transforms of the list, the method comprises a step of selecting a transform from among the transforms of the list applied to the current block according to a distortion rate criterion and a step of coding information representative of a position index of the transform selected in the list obtained. 9. Method for decoding a digital image according to one of claims 2 to 7, characterized in that, when the list includes several transforms, the step of identifying the transform further comprises a step of decoding information identifying the transform in the list. 10. Device (100) for coding a digital image, said image (Ik) being divided into a plurality of blocks of pixels processed in a defined order, said device comprising a calculation machine dedicated to and configured for: Predict (PRED) values of the current block from at least one block previously processed from a previous or next image, called the reference image, Calculate (CALC) a residual block (R) by subtracting the predicted values from the original values of the current block, Transform (TRANSF) the residual block by applying a transform; Code (COD) the transformed residue block; characterized in that the calculating machine is further configured to obtain (OBT IV) an indicator of variability of the current block with respect to said at least one reference image, to obtain (OBT L) a list of transforms among a plurality of lists predetermined at least as a function of the indicator obtained and to apply to the residue block a transform identified in the list obtained. 11. Device (200) for decoding an image of a sequence of images from coded data representative of said image, said image being divided into a plurality of blocks processed in a defined order, said device comprising a machine for calculation dedicated to and configured for: Predict (PRED) a current block from at least one previously processed block of a previous or next image, called the reference image; Decode (DEC) coded values of the coefficients of the transformed residual block extracted from the binary train; Carry out a reverse transformation (TRANS-1) of the transformed residue block; Reconstruct (RECONST) the decoded block from the residual block and from the prediction of the current block; said device being characterized in that the calculating machine is further configured to obtain (OBT IV) an indicator of variability for predicting the current block with respect to at least one reference image, obtaining (OBT L) a list of transforms from a plurality of predetermined lists at least as a function of the indicator obtained and identifying (ID TR) the transformed in the list obtained. 12. Signal carrying a bit stream (TB) comprising coded data representative of an image of a sequence of images, said image being divided into blocks of pixels processed in a defined order, the values of a current block being predicted from the values of at least one block previously processed from at least one previous or next image, called the reference image, the values 5 of a residual block being calculated by subtracting the predicted values from the original values of the current block, a transformed residual block being obtained by applying a transform to pixels of the residual block; characterized in that said signal comprises coded information representative of a position index of the transform to be applied to the current residual block in a list comprising a 10 ordered sequence of transforms, the list being intended to be obtained at least as a function of an indicator of variability of the current block with respect to said at least one reference image. 13. Terminal equipment (TU) characterized in that it comprises a transmission module (E / R) and a module (E / R) for receiving data to and from a telecommunications network, a device ( 100) for coding a digital image according to claim 10 and a device 15 decoding (200) according to claim 11. 14. Computer program (Pgl) comprising instructions for implementing the method of coding a digital image according to one of claims 1 and 3 to 8, when executed by a processor. 15. Computer program comprising instructions for implementing the method of decoding a digital image according to one of claims 2 to 7 and 9, when executed by a processor. 1/8