EP2724536A1

EP2724536A1 - Method for encoding and decoding images, encoding and decoding device, and corresponding computer programs

Info

Publication number: EP2724536A1
Application number: EP12734974.4A
Authority: EP
Inventors: Felix Henry; Stéphane PATEUX; Gordon Clare
Original assignee: Orange SA
Current assignee: Dolby International AB
Priority date: 2011-06-24
Filing date: 2012-06-20
Publication date: 2014-04-30
Also published as: KR102095329B1; BR112013032956A2; KR20160086984A; RU2018103979A; KR20200135580A; CN103959787A; RU2018103979A3; US9380308B2; US20170223353A1; HK1244374A1; JP2017073806A; US9848196B2; US20140254665A1; US20150195537A1; CN107094255A; JP7280233B2; KR102437249B1; US9319694B2; US20160234512A1; US10362311B2

Abstract

The invention relates to an encoding method including cutting out (C1) the image from a plurality of blocks (MB) capable of containing symbols belonging to a predetermined set of symbols, grouping (C2) blocks in a predetermined number (P) of subsets of blocks (SE1, SE2,…SEk,…,SEP), encoding (C3), using an entropic module, each of said subsets of blocks, by combining digital information with the symbols of each block of a subset in question, said encoding step including, for the first block of the image, a sub-step (C33) of initializing status variables of said entropic encoding module, and then generating at least one sub-stream of data representative of at least one of said subsets of encoded blocks. In the event the current block is the first block being encoded of a subset of blocks in question, the probabilities of a symbol appearing for said first current block are those determined for a predetermined encoded and decoded block of at least one other subset. In the event the current unit is the last encoded unit of the subset in question, all of the digital information that was associated with the symbols during the encoding of the blocks of the said subset in question are written (C45) in the sub-stream representative of at least said subset in question, and the initialization sub-step is implemented (C46).

Description

METHOD FOR ENCODING AND DECODING IMAGES, DEVICE FOR ENCODING AND DECODING AND COMPUTER PROGRAMS

CORRESPONDENTS Field of the invention

The present invention relates generally to the field of image processing, and more specifically to the encoding and decoding of digital images and digital image sequences.

The invention can thus, in particular, apply to the video coding implemented in current (MPEG, H.264, etc.) or future video encoders (ITU-T / VCEG (H.265) or ISO / MPEG ( HVC).

Background of the invention

Current video encoders (MPEG, H264, ...) use a block representation of the video sequence. The images are divided into macroblocks, each macroblock is itself divided into blocks and each block, or macroblock, is coded by intra-image prediction or inter-image prediction. Thus, some images are coded by spatial prediction (intra prediction), while other images are coded by temporal prediction (inter prediction) with respect to one or more coded-decoded reference images, by means of compensation. in motion known to those skilled in the art. In addition, for each block can be coded a residual block corresponding to the original block minus a prediction. The coefficients of this block are quantized after a possible transformation, then coded by an entropic coder.

Intra prediction and inter prediction require that some blocks that have been previously coded and decoded be available, so as to be used, both at the decoder and the encoder, to predict the current block. A schematic example of such a predictive coding is shown in FIG. 1, in which an image I _N is divided into blocks, a current block MB of this image being subjected to a predictive coding with respect to a predetermined number of three blocks. MBr-ι, MBr ₂ and MBr ₃ previously coded and decoded, as indicated by the gray arrows. The three aforementioned blocks specifically include the MBn block located immediately to the left of the current block MB ,, and the two blocks MBr ₂ and MBr ₃ respectively located immediately above and to the right above the current block MB ,.

We are interested here more particularly in the entropic coder. The entropic encoder encodes the information according to their order of arrival. Typically a line-by-line path of the blocks is made, of "raster-scan" type, as illustrated in FIG. 1 by the reference PRS, starting from the block at the top left of the image. For each block, the various information necessary for the representation of the block (block type, prediction mode, residual coefficients, ...) are sent sequentially to the entropy coder.

There is already known an efficient arithmetic coder of reasonable complexity, called CABAC (Context Adaptive Binary Arithmetic Coder), introduced in the AVC compression standard (also known as ISO-MPEG4 Part 10 and ITU-A). T H.264).

This entropic coder implements different concepts:

arithmetic coding: the coder, as initially described in J. Rissanen and G. G. Langdon, "Universal modeling and coding," IEEE Trans. Inform. Theory, vol. IT-27, pp. 12-23, Jan. 1981, uses, to encode a symbol, a probability of appearance of this symbol;

- Adaptation to the context: it is a question of adapting the probability of appearance of the symbols to be coded. On the one hand, an on-the-fly learning is done. On the other hand, depending on the state of the previously coded information, a specific context is used for the coding. Each context corresponds to a probability of appearance of the symbol itself. For example, a context corresponds to a type of coded symbol (the representation of a coefficient of a residue, coding mode signaling, etc.) according to a given configuration, or a neighborhood state (for example the number of modes). "Intra" selected in the neighborhood, ...);

binarization: a setting in the form of a sequence of bits of the symbols to be encoded is performed. Subsequently, these different bits are successively sent to the binary entropic coder.

Thus, this entropic coder implements, for each context used, a system for learning on-the-fly probabilities with respect to symbols previously coded for the context in question. This learning is based on the coding order of these symbols. Typically, the image is scanned according to a "raster-scan" type order, described above.

When coding a given symbol b that can be 0 or 1, the learning of the probability p of appearance of this symbol is updated for a current block MB, as follows: where a is a predetermined value, for example 0.95 and p _M is the probability of appearance of symbol calculated at the last appearance of this symbol.

A schematic example of such an entropy coding is shown in FIG. 1, in which a current block MB of the image 1 _N is subjected to entropy coding. When the entropic coding of the block MB begins, the probabilities of appearance of symbols used are those obtained after coding a previously coded and decoded block, which is the one immediately preceding the current block MB, in accordance with the line-by-line blocks of the type "raster scan" supra. Such learning based block dependency is shown in Figure 1 for some blocks only for the sake of clarity of the figure, by the arrows fine line.

A disadvantage of such a type of entropy encoding lies in the fact that when encoding a symbol at the beginning of a line, the probabilities used correspond mainly to those observed for the symbols at the end of the previous line. , taking into account the "raster scan" path of the blocks. However, because of the possible spatial variation of the probabilities of the symbols (for example for a symbol linked to a movement information, the movement situated on the right part of an image may be different from that observed on the left part and therefore of even for the resulting local probabilities), a lack of local matching of the probabilities can be observed, which may lead to a loss of efficiency during coding. To limit this phenomenon, proposals for changes to the order of the blocks have been made, in order to ensure better local consistency, but the coding and decoding remain sequential.

This is another disadvantage of this type of entropic coder. Indeed, the coding and the decoding of a symbol being dependent on the state of the probability learned so far, the decoding of the symbols can be done in the same order as that used during the coding. Typically, the decoding can then be only sequential, thus preventing a parallel decoding of several symbols (for example to take advantage of multi-core architectures).

The document: Thomas Wiegand, Gary J. Sullivan, Gisle Bjontegaard, and Ajay Luthra, "Overview of the H.264 / Video Coding Standard", IEEE Transactions on Circuits and Systems for Video Technology, Vol. 13, No. 7, pp. 560-576, July 2003, further specifies that the entropic coder CABAC has the particularity of assigning a non-integer number of bits to each symbol of a current alphabet to be encoded, which is advantageous for the probabilities of appearance of symbols. greater than 0.5. Specifically, the CABAC encoder waits to read several symbols, then assigns to this set of symbols read a predetermined number of bits that the encoder registered in the compressed stream to be transmitted to the decoder. Such an arrangement thus makes it possible to "mutualize" the bits over several symbols and to encode a symbol on a fractional number of bits, this number reflecting information that is closer to the information actually conveyed by a symbol. Other bits associated with the symbols read are not transmitted in the compressed stream but are kept waiting to be assigned to one or more new symbols read by the CABAC coder again to mutualize these other bits. In a manner known per se, the entropic coder proceeds at a given instant to "emptying" these non-transmitted bits. In other words, at said given instant, the encoder extracts the bits not yet transmitted and writes them in the compressed stream to the decoder. Such emptying occurs for example at the moment when the last symbol to be coded has been read, so as to ensure that the compressed stream contains all the bits that will allow the decoder to decode all the symbols of the alphabet. In a general view, the time at which the drain is performed is determined by the performance and functionality of a particular encoder / decoder.

The document, which is available at http://research.microsoft.com/en-us/um/people/iinl/paper 2002 / msri jpeg.htm as of April 15, 201 1, describes a process still image encoding compliant with the JPEG2000 compression standard. In this method, the still image data undergo discrete wavelet transform followed by quantization, thereby obtaining quantized wavelet coefficients to which quantization indices are respectively associated. Quantization indices obtained are encoded using an entropy coder. The quantized coefficients are previously grouped in rectangular blocks called codeblocks, typically of size 64x64 or 32x32. Each codeblock is then independently encoded by entropy coding. Thus, the entropic coder, when coding a current block code, does not use the appearance probabilities of symbols calculated during the coding of previous codeblocks. The entropic coder is therefore in a state initialized at the beginning of coding a code-block. Such a method has the advantage of decoding the data of a code-block without having to decode neighboring code blocks. For example, client software may require server software to provide compressed codeblocks that the client only needs to decode an identified sub-part of an image. Such a method also has the advantage of authorizing the encoding and / or the decoding in parallel of the code blocks. Thus, the smaller the code-blocks, the higher the level of parallelism. For example, for a level of parallelism set at two, two codeblocks will be encoded and / or decoded in parallel. In theory, the value of the level of parallelism is equal to the number of code blocks to be encoded in the image. However, the compression performance obtained with this method is not optimal considering that such a coding does not take advantage of probabilities arising from the immediate environment of the current code-block. Object and summary of the invention

One of the aims of the invention is to overcome disadvantages of the state of the art mentioned above.

For this purpose, an object of the present invention relates to a method of encoding at least one image comprising the steps of:

- cutting the image into a plurality of blocks able to contain symbols belonging to a predetermined set of symbols,

grouping blocks into a predetermined number of subsets of blocks,

coding, by means of an entropy coding module, of each of the subsets of blocks, by association of digital information with the symbols of each block of a subset considered, the coding step comprising, for the first block of the image, a substep of initialization of state variables of the entropy coding module,

generating at least one sub-data stream representative of at least one of the coded block subsets.

The process according to the invention is remarkable in that:

in the case where the current block is the first block to code a subset considered, it is proceeded to the determination of symbol occurrence probabilities for the first current block, the probabilities being those which have been determined for a predetermined coded and decoded block of at least one other subset,

in the case where the current block is the last coded block of the subset considered, it is proceeded to:

· Writing, in the representative sub-stream of the subset considered, all the digital information that has been associated with the symbols during the coding of the blocks of the subset considered,

• the implementation of the initialization sub-step.

The writing step mentioned above amounts to carrying out, as soon as the last block of a subset of blocks has been coded, an emptying of the digital information (bits) not yet transmitted, as explained above. in the description. Coupling the aforementioned write step and the entropy coding module reset step produces an encoded data stream containing different data sub-streams corresponding to at least one coded block subset, respectively. said stream being adapted to be decoded in parallel according to different levels of parallelism, regardless of the type of coding, sequential or parallel, which has been applied to the subsets of blocks. Thus, a large degree of freedom can be obtained at decoding on the choice of the level of parallelism, according to the expected coding / decoding performance. The level of parallelism to the decoding is variable and may even be different from the level of parallelism with the coding, since at the start of the decoding of a subset of blocks, the decoder is always in an initialized state.

According to a first example, the state variables of the entropy coding module are the two terminals of an interval representative of the probability of appearance of a symbol among the symbols of the predetermined set of symbols.

According to a second example, the state variables of the entropy coding module are the symbol strings contained in the translation table of a LZW entrapment coder (Lempel-Ziv-Welch) well known to those skilled in the art, and described at the following internet address as of June 21, 201 1: http: //en.wikipedia.orq/wiki/lempel%E2%80%93Ziv%E2%80%93Weich.

The use of the symbol occurrence probabilities determined for the first block of said other subset during the entropy coding of the first current block of a subset of blocks considered mainly has the advantage of saving the buffer memory of the encoder by storing in the latter only updating said probabilities of appearance of symbols, without taking into account the probabilities of appearance of symbols learned by the other consecutive blocks of said other subset.

The use of the symbol occurrence probabilities determined for a block of said other subset, other than the first block, for example the second block, during the entropic coding of the first current block of a subset of blocks considered a mainly for the benefit of obtaining a more accurate learning and therefore better probability of occurrence of symbols, resulting in better video compression performance.

In a particular embodiment, the subsets of blocks are coded sequentially or in parallel.

The fact that the block subsets are coded sequentially has the advantage of making the coding method according to the invention in accordance with the H.264 / MPEG-4 AVC standard.

The fact that the block subsets are coded in parallel has the advantage of accelerating the encoder processing time and benefiting from a multiplatform architecture for coding an image.

In another particular embodiment, when at least two subsets of blocks are coded parallel to at least one other subset of blocks, the at least two subsets of coded blocks are contained in the same sub-stream. of data.

Such an arrangement makes it possible in particular to save on the signaling of the sub-data streams. In fact, for a decoding unit to be able to decode a sub-stream as soon as possible, it is necessary to indicate in the compressed file where the sub-stream in question begins. When multiple subsets of blocks are contained in the same data sub-stream, only one flag is needed, which reduces the size of the compressed file.

In yet another particular embodiment, when the subsets of coded blocks are intended to be decoded in parallel according to a predetermined order, the sub-data streams delivered after coding respectively of each of the subsets of blocks are previously ordered. in the predetermined order before being transmitted for decoding.

Such an arrangement makes it possible to adapt the coded data stream to a specific type of decoding without the need to decode and re-encode the image.

Correlatively, the invention also relates to a device for coding at least one image comprising: means for cutting the image into a plurality of blocks able to contain symbols belonging to a predetermined set of symbols,

means for grouping the blocks into a predetermined number of subsets of blocks,

coding means for each of the subsets of blocks, the coding means comprising an entropy coding module able to associate digital information with the symbols of each block of a subset considered, the coding means comprising, for the first block of the image, means of initialization of state variables of the entropy coding module,

means for generating at least one data sub-stream representative of at least one of the coded block subsets.

Such a coding device is remarkable in that it comprises:

means for determining symbol appearance probabilities for a current block which, in the case where the current block is the first block to be encoded for a considered subset, determine the symbol appearance probabilities for the first block block as those which have been determined for a predetermined coded and decoded block of at least one other subset,

writing means which, in the case where the current block is the last coded block of the subset considered, are activated to write, in the sub-stream representative of the subset considered, all the digital information which has have been associated with the symbols during the coding of the blocks of the subset considered,

the initialization sub-means being further activated to reset the state variables of the entropy coding module.

Correspondingly, the invention also relates to a method for decoding a representative stream of at least one coded picture, comprising the steps of:

identification in the stream of a predetermined number of sub-data streams respectively corresponding to at least one subset of blocks to be decoded, the blocks being able to contain symbols belonging to a predetermined set of symbols,

decoding the subsets of blocks identified by means of an entropy decoding module, by reading, in at least one of the identified substreams, digital information associated with the symbols of each block of the subset corresponding to the at least one an identified substream, the decoding step comprising, for the first block to be decoded from the image, a substep of initialization of state variables of the entropy decoding module.

Such a decoding method is remarkable in that:

in the case where the current block is the first block to be decoded from a subset considered, it is proceeded to the determination of symbol occurrence probabilities for the first block of the subset considered, the probabilities being those which have been determined for a predetermined decoded block of at least one other subset,

- In the case where the current block is the last decoded block of the subset considered, it is proceeded to the implementation of the initialization sub-step.

In a particular embodiment, the subsets of blocks are decoded sequentially or in parallel.

In another particular embodiment, when at least two subsets of blocks are decoded in parallel with at least one other subset of blocks, one of the identified data substreams is representative of the at least two subsets of blocks.

In yet another particular embodiment, when the subsets of coded blocks are intended to be decoded in parallel in a predetermined order, the sub-data streams respectively corresponding to the subsets of coded blocks are previously ordered according to said order predetermined in said stream to be decoded.

Correlatively, the invention further relates to a device for decoding a representative stream of at least one coded picture, comprising:

means for identifying in the stream a predetermined number of data sub-streams respectively corresponding to the minus a subset of blocks to be decoded, the blocks being able to contain symbols belonging to a predetermined set of symbols,

means for decoding the subsets of identified blocks, the decoding means comprising an entropy decoding module capable of reading, in at least one of the identified subflows, digital information associated with the symbols of each block of the subset; corresponding to said at least one identified sub-stream, the decoding means comprising, for the first block to be decoded from the image, initialization sub-means of state variables of the entropy decoding module.

Such a decoding device is remarkable in that it comprises means for determining probabilities of symbol appearance for a current block which, in the case where the current block is the first block to be decoded of a considered subset. , determine the symbol occurrence probabilities for the first block as those determined for a decoded predetermined block of at least one other subset,

and in that in the case where the current block is the last decoded block of the subset considered, the initialization sub-means are activated to reset the state variables of the entropy decoding module.

The invention also relates to a computer program comprising instructions for executing the steps of the above coding or decoding method, when the program is executed by a computer.

Such a program can use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in a partially compiled form, or in any another desirable form.

Still another object of the invention is directed to a computer readable recording medium, and including computer program instructions as mentioned above.

The recording medium may be any entity or device capable of storing the program. For example, such a medium may comprise storage means, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or a magnetic recording medium, for example a floppy disk or a Hard disk. On the other hand, such a recording medium may be a transmissible medium such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by other means. The program according to the invention can be downloaded in particular on an Internet type network.

Alternatively, such a recording medium may be an integrated circuit in which the program is incorporated, the circuit being adapted to execute the method in question or to be used in the execution of the latter.

The coding device, the decoding method, the decoding device and the aforementioned computer programs have at least the same advantages as those conferred by the coding method according to the present invention. Brief description of the drawings

Other features and advantages will appear on reading two preferred embodiments described with reference to the figures in which:

FIG. 1 represents an image coding scheme of the prior art,

FIG. 2A represents the main steps of the coding method according to the invention,

FIG. 2B shows in detail the coding implemented in the coding method of FIG. 2A,

FIG. 3A represents a first embodiment of a coding device according to the invention,

FIG. 3B represents a coding unit of the coding device of FIG. 3A,

FIG. 3C represents a second embodiment of a coding device according to the invention,

FIG. 4A represents an image coding / decoding scheme according to a first preferred embodiment,

FIG. 4B represents an image coding / decoding scheme according to a second preferred embodiment, FIG. 5A represents the main steps of the decoding method according to the invention,

FIG. 5B shows in detail the decoding implemented in the decoding method of FIG. 5A,

FIG. 6A represents an embodiment of a decoding device according to the invention,

FIG. 6B represents a decoding unit of the decoding device of FIG. 6A,

FIG. 7A represents an image coding / decoding scheme implementing a sequential type coding and a parallel type decoding,

FIG. 7B represents an image coding / decoding scheme implementing a parallel type coding / decoding, with different levels of parallelism respectively.

Detailed description of a first embodiment of the coding part

An embodiment of the invention will now be described, in which the coding method according to the invention is used to code a sequence of images according to a bit stream close to that obtained by a coding according to the H standard. .264 / MPEG-4 AVC. In this embodiment, the coding method according to the invention is for example implemented in a software or hardware way by modifications of an encoder initially compliant with the H.264 / MPEG-4 AVC standard. The coding method according to the invention is represented in the form of an algorithm comprising steps C1 to C5, represented in FIG. 2A.

According to the embodiment of the invention, the coding method according to the invention is implemented in a coding device CO, two embodiments of which are respectively represented in FIGS. 3A and 3C.

With reference to FIG. 2A, the first coding step C1 is the division of an image IE of an image sequence to be encoded into a plurality of blocks or macroblocks MB, as represented in FIG. 4A or 4B. Said macroblocks are capable of containing one or more symbols, said symbols forming part of a predetermined set of symbols. In the examples shown, said MB blocks have a square shape and are all the same size. Depending on the size of the image that is not necessarily a multiple of the size of the blocks, the last blocks on the left and the last blocks on the bottom may not be square. In an alternative embodiment, the blocks may be for example of rectangular size and / or not aligned with each other.

Each block or macroblock can also be itself divided into sub-blocks that are themselves subdividable.

Such splitting is performed by a partitioning PCO module shown in FIG. 3A which uses, for example, a partitioning algorithm that is well known as such.

With reference to FIG. 2A, the second coding step C2 is the grouping of the aforementioned blocks into a predetermined number P of consecutive subsets of blocks SE1, SE2,..., SEk,..., SEP intended to be coded sequentially. or in parallel. In the examples shown in FIGS. 4A and 4B, P = 6, but only four subsets SE1, SE2, SE3, SE4 are shown for the sake of clarity of the figures. These four subsets of blocks are each represented in dashed line and consist respectively of the first four block lines of the image IE.

Such a grouping is performed by a GRCO calculation module represented in FIG. 3A, using a well-known algorithm per se.

With reference to FIG. 2A, the third coding step C3 consists in the coding of each of said subsets of blocks SE1 to SE6, the blocks of a subset considered being coded according to a predetermined PS travel order, which is for example of sequential type. In the examples shown in Figures 4A and 4B, the blocks of a subset SEk current (1≤k≤P) are coded one after the other, from left to right, as indicated by the arrow PS.

According to a first variant, such an encoding is of the sequential type and is implemented by a single coding unit UC as represented in FIG. 3A. In a manner known per se, the coder CO comprises a buffer MT which is adapted to contain the probability of appearance of symbols such as progressively updated as and when encoding a current block.

As shown in more detail in FIG. 3B, the coding unit UC comprises:

A coding module predictive of a current block with respect to at least one previously coded and decoded block, denoted MCP;

An entropy coding module of said current block by using at least one symbol appearance probability calculated for said previously coded and decoded block, denoted ECM.

The predictive coding module MCP is a software module that is able to carry out a predictive coding of the current block, according to conventional prediction techniques, such as for example in Intra and / or Inter mode.

The entropic coding module MCE is CABAC type, but modified according to the present invention as will be described later in the description.

Alternatively, the entropic coding module MCE could be a Huffman coder known as such.

In the examples shown in FIGS. 4A and 4B, the unit UC encodes the blocks of the first line SE1, from left to right. When it reaches the last block of the first line SE1, it goes to the first block of the second line SE2. When it reaches the last block of the second line SE2, it goes to the first block of the third line SE3. When it reaches the last block of the third line SE3, it goes to the first block of the fourth line SE4, and so on until the last block of the image IE is coded.

Other types of course than the one just described above are of course possible. Thus, it is possible to cut the IE image into several sub-images and to independently apply a division of this type on each sub-image. It is also possible for the coding unit to process not a succession of lines, as explained above, but a succession of columns. It is also possible to browse the rows or columns in one direction or the other.

According to a second variant, such an encoding is of the parallel type and differs from the first variant of sequential coding only in that it is implemented by a predetermined number R of coding units UCk (1 k k cod R), with R = 2 in the example shown in Figure 3C. Such parallel coding is known to generate a substantial acceleration of the coding method.

Each of the coding units UCk is identical to the coding unit UC shown in FIG. 3B. Correspondingly, a coding unit UCk comprises a predictive coding module MCPk and an entropy coding module MCEk.

Referring again to FIGS. 4A and 4B, the first unit UC1 codes, for example, the blocks of lines of odd rank, whereas the second unit UC2 codes, for example, the blocks of lines of even rank. Specifically, the first unit UC1 codes the blocks of the first line SE1, from left to right. When it reaches the last block of the first line SE1, it goes to the first block of the (2n + 1) ^th line, that is to say the third line SE3, etc. In parallel with the processing performed by the first unit UC1, the second unit UC2 codes the blocks of the second line SE2, from left to right. When it reaches the last block of the second line SE2, it goes to the first block of the (2n) ^th line, here the fourth line SE4, etc. The two aforementioned paths are repeated until the last block of the IE image is encoded.

With reference to FIG. 2A, the fourth encoding step C4 is the production of L sub-flux F1, F2,..., Fm,..., FL (1 <m <L <P) of bits representing the blocks processed by the aforementioned UC encoding unit or each of the aforementioned UCk coding units, as well as a decoded version of the processed blocks of each subset SEk. The decoded processed blocks of a considered subset, denoted SED1, SED2, ..., SEDk, ..., SEDP are likely to be reused by the coding unit UC represented in FIG. 3A or each encoding units UCk shown in Figure 3C, according to a synchronization mechanism which will be detailed later in the description.

With reference to FIG. 3B, the sub-stream production step L is implemented by a flow generating software module MGSF or MGSFk which is adapted to produce data streams, such as bits for example.

With reference to FIG. 2A, the fifth encoding step C5 consists in constructing a global flux F from the aforementioned L subflows F1, F2,..., Fm,... FL. According to one embodiment, the substreams F1, F2,..., Fm,..., FL are simply juxtaposed, with additional information intended to indicate to the decoder the location of each sub-flow Fm in the stream. global F. The latter is then transmitted by a communication network (not shown) to a remote terminal. This includes the decoder DO shown in FIG. 5A. According to another particularly advantageous embodiment because it does not require decoding and then re-encoding the image, the coder CO, before transmitting the stream F to the decoder DO, previously orders the L substreams F1, F2, ... , Fm, ..., FL according to a predetermined order which corresponds to the order in which the decoder DO is able to decode the sub-streams.

Thus, as will be described in detail later in the description, the decoder according to the invention is able to isolate the substreams F1, F2, ..., Fm, ..., FL within the overall flow F and assign them to one or more decoding units composing the decoder. It will be noted that such a decomposition of the subflows into a global stream is independent of the choice of the use of a single coding unit or of several coding units operating in parallel, and that it is possible with this approach to only have the encoder or only the decoder which includes units operating in parallel.

Such a construction of the global flow F is implemented in a flow generating module CF, as shown in FIG. 3A and FIG. 3C.

Reference will now be made to FIG. 2B, the different specific sub-steps of the invention, as implemented during the aforementioned coding step C3, in a UC or UCk coding unit. During a step C31, the encoding unit UC or UCk selects as the current block the first block to be encoded of a current line SEk shown in FIG. 4A or 4B, such as for example the first line SE1.

During a step C32, the unit UC or UCk tests whether the current block is the first block (located at the top left) of the image IE which has been cut into blocks in the aforementioned step C1.

If this is the case, during a step C33, the entropic coding module MCE or MCEk initializes its state variables. According to the example shown, which uses the arithmetic coding described above, it is an initialization of an interval representative of the probability of appearance of a symbol contained in the predetermined set of symbols. In known manner as such, this interval is initialized with two terminals L and H, respectively lower and upper. The value of the lower bound L is set to 0, while the value of the upper bound is set to 1, which corresponds to the probability of occurrence of a first symbol among all the symbols of the predetermined set of symbols. . The size R of this interval is therefore defined at this stage by R = H - L = 1. The initialized interval is further conventionally partitioned into a plurality of predetermined sub-intervals which are respectively representative of the appearance probabilities of the symbols of the predetermined set of symbols.

Alternatively, if the entropy encoding used is LZW encoding, a symbol string translation table is initialized, so that it contains all possible symbols once and only once.

If following the aforementioned step C32, the current block is not the first block of the image IE, it is proceeded during a step C40 which will be described later in the following description, to determining the availability of the previously coded and decoded blocks needed.

During a step C34, the first current block MB1 of the first line SE1 shown in FIG. 4A or 4B is coded. Such a step C34 comprises a plurality of sub-steps C341 to C348 which will be described below. During a first substep C341 represented in FIG. 2B, the predictive coding of the current block MB1 is carried out by known intra and / or inter prediction techniques, during which the block MB1 is predicted with respect to least one block previously coded and decoded.

It goes without saying that other intra prediction modes as proposed in the H.264 standard are possible.

The current block MB1 may also be subjected to prediction coding in inter mode, during which the current block is predicted with respect to a block resulting from a previously coded and decoded picture. Other types of prediction are of course conceivable. Among the possible predictions for a current block, the optimal prediction is chosen according to a distortion flow criterion well known to those skilled in the art.

Said aforementioned predictive coding step makes it possible to construct a predicted block MBp-ι which is an approximation of the current block MB- ,. The information relating to this predictive coding will subsequently be written in the stream F transmitted to the decoder DO. Such information includes in particular the type of prediction (inter or intra), and if appropriate, the intra prediction mode, the type of partitioning of a block or macroblock if the latter has been subdivided, the image index of reference and displacement vector used in the inter prediction mode. This information is compressed by the CO encoder.

During a subsequent substep C342, the predicted block MBpi of the current block MBi is subtracted to produce a residue block MBr-i.

During a subsequent substep C343, the residue block ΜΒη is transformed according to a conventional direct transformation operation such as, for example, a discrete cosine transformation of the DCT type, to produce a transformed block MBt-i.

During a subsequent substep C344, the transformed block MBti is quantized according to a conventional quantization operation, such as, for example, a scalar quantization. A block of quantized coefficients MBqi is then obtained. During a subsequent sub-step C345, the entropic coding of the quantized coefficient block MBq-, is carried out. In the preferred embodiment, it is a CABAC entropic coding. Such a step consists of:

a) reading the symbol or symbols of the predetermined set of symbols which are associated with said current block,

b) associating digital information, such as bits, with the read symbol (s).

In the aforementioned variant according to which the coding used is an LZW coding, digital information corresponding to the code of the symbol in the current translation table is associated with the symbol to be coded, and an update of the translation table is carried out, according to a method known per se.

During a subsequent substep C346, the block MBq is dequantized according to a conventional dequantization operation, which is the inverse operation of the quantization performed in step C344. A block of dequantized coefficients MBDq is then obtained.

During a subsequent substep C347, the inverse transformation of the dequantized coefficient block MBDqi is carried out which is the inverse operation of the direct transformation carried out in step C343 above. A decoded residue block MBDr-ι is then obtained.

During a subsequent substep C348, the decoded block MBDi is constructed by adding or prediction block MBpi the decoded residue block MBDr-i. It should be noted that this last block is the same as the decoded block obtained at the end of the decoding process of the IE image which will be described later in the description. The decoded block MBDi is thus made available for use by the coding unit UCk or any other coding unit forming part of the predetermined number R of coding units.

At the end of the aforementioned coding step C34, the entropic coding module MCE or MCEk as represented in FIG. 3B contains all the probabilities such as progressively being updated as the first block is coded. These probabilities correspond to the different elements of possible syntaxes and to the different coding contexts associated. Following the aforementioned coding step C34, it is tested, during a step C35, if the current block is the jth block of this same line, where j is a known predetermined value of the CO encoder which is at less than 1.

If this is the case, during a step C36 represented in FIG. 2B, the set of probabilities calculated for the jth block is stored in the buffer memory MT of the coder CO as represented in FIG. 3A or 3B and FIGS. FIGS. 4A and 4B, the size of said memory being adapted to store the number of calculated probabilities.

During a step C37 represented in FIG. 2B, the coding unit UC or UCk tests whether the current block of the line SEk that has just been encoded is the last block of the image IE. Such a step is also implemented if during step C35, the current block is not the jth block of the line SE1.

If the current block is the last block of the image IE, during a step C38, the coding process is terminated.

If this is not the case, it is proceeded, in step C39, to the selection of the next block MB, to be encoded according to the order of travel represented by the arrow PS in Figure 4A or 4B.

During a step C40 shown in FIG. 2B, the availability of previously coded and decoded blocks which are necessary to code the current block MB, is determined.

If it is the first line SE1, such a step consists in checking the availability of at least one block to the left of the current block to be coded MB ,. However, given the PS travel order chosen in the embodiment shown in Figure 4A or 4B, the blocks are encoded one after the other on a line SEk considered. As a result, the coded and decoded block on the left is always available (except for the first block of a line). In the example shown in FIG. 4A or 4B, it is the block located immediately to the left of the current block to be coded.

If it is a line SEk different from the first line, said determining step further comprises checking whether a predetermined number N 'of blocks located on the previous line SEk-1, for example the two blocks located respectively above and to the right of the current block, are available for the coding of the current block, that is to say if they have already been coded and then decoded by the UC or UCk-1 coding unit.

This test step being able to slow down the coding method, alternatively according to the invention, in the case where the coding of the lines is of parallel type, a clock CLK shown in FIG. 3C is adapted to synchronize the advanced block coding so as to guarantee the availability of the two blocks located respectively above and above right of the current block, without it being necessary to check the availability of these two blocks. Thus, a coding unit UCk always starts coding the first block with an offset of a predetermined number N '(with for example N' = 2) of coded and decoded blocks of the preceding line SEk-1 which are used for the coding of the current block. From a software point of view, the implementation of such a clock makes it possible to significantly accelerate the processing time of the blocks of the image IE in the coder CO.

During a step C41 shown in Figure 2B, it is tested if the current block is the first block of the SEk line considered.

If this is the case, during a step C42, only the probabilities of appearance of symbols calculated during the coding of the jth block of the preceding line SEk-1 are read in the buffer memory MT.

According to a first variant represented in FIG. 4A, the jth block is the first block of the preceding line SEk-1 (j = 1). Such a reading consists in replacing the probabilities of the CABAC encoder by those present in the buffer memory MT. With regard to the first respective blocks of the second, third and fourth lines SE2, SE3 and SE4, this reading step is shown in FIG. 4A by the arrows represented in fine lines.

According to a second variant of the above-mentioned step C43 which is illustrated in FIG. 4B, the jth block is the second block of the previous line SEk-1 (j = 2). Such a reading consists in replacing the probabilities of the CABAC encoder by those present in the buffer memory MT. With regard to the first respective blocks of the second, third and fourth lines SE2, SE3 and SE4, this reading step is shown in FIG. 4B by the arrows represented in dashed fine lines. Following step C42, the current block is coded and then decoded by iteration of the steps C34 to C38 described above.

If following the aforementioned step C41, the current block is not the first block of the line SEk considered, it is advantageously not proceeded to the reading of the probabilities from the previously coded and decoded block which is on the same line SEk, that is to say the encoded and decoded block located immediately to the left of the current block, in the example shown. In fact, given the sequential reading path PS of the blocks located on the same line, as represented in FIG. 4A or 4B, the probabilities of appearance of symbols present in the CABAC coder at the time of the beginning of the coding of the current block are exactly those which are present after coding / decoding of the previous block on this same line.

Consequently, during a step C43 represented in FIG. 2B, the symbol occurrence probabilities for the entropic coding of said current block are learned, which only correspond to those calculated for said block. previous in the same line, as represented by the double arrows in full lines in Figure 4A or 4B.

Following step C43, the current block is coded and then decoded by iteration of the steps C34 to C38 described above.

It is then tested, during a step C44, if the current block is the last block of the line SEk considered.

If this is not the case, following step C44, step C39 of selecting the next block MB to be coded is again implemented.

If the current block is the last block of the line SEk considered, during a step C45, the coding device CO of Figure 3A or 3C performs a drain as mentioned above in the description. For this purpose, the coding unit UCk transmits to the corresponding sub-stream generation module MGSFk all the bits that have been associated with the symbol (s) read during the coding of each block of said line. SEk considered, so that the MGSFk module writes, in the sub-data stream Fm containing a bitstream representative of the coded blocks of said line SEk considered, said totality of bits. Such emptying is symbolized in FIGS. 4A and 4B by a triangle at the end of each line SEk. During a step C46 represented in FIG. 2B, the coding unit UC or UCk performs a step identical to the aforementioned step C33, that is to say, it initializes again the interval representative of the probability of appearance of a symbol contained in the predetermined set of symbols. Such a reset is shown in Figures 4A and 4B by a black dot at the beginning of each line SEk.

The advantage of performing steps C45 and C46 at this level of the coding is that during the coding of the next block processed by the coding unit UC or a coding unit UCk, the coder CO is in an initialized state. Thus, as will be described later in the description, it becomes possible for a decoding unit working in parallel to directly decode the compressed stream F from that point, since it suffices to be in the initialized state.

Detailed description of an embodiment of the decoding part

An embodiment of the decoding method according to the invention will now be described, in which the decoding method is implemented in a software or hardware way by modifying a decoder initially compliant with the H.264 / MPEG-4 AVC standard.

The decoding method according to the invention is represented in the form of an algorithm comprising steps D1 to D4, represented in FIG. 5A.

According to the embodiment of the invention, the decoding method according to the invention is implemented in a decoding device DO shown in FIG. 6A.

With reference to FIG. 5A, the first decoding step D1 is the identification in said stream F of the L substreams F1, F2,..., Fm,..., FL respectively containing the P subsets SE1, SE2, ..., SEk, ..., SEP of previously coded blocks or macroblocks MB, as shown in FIG. 4A or 4B. For this purpose, each sub-flow Fm in the stream F is associated with an indicator intended to allow the decoder DO to determine the location of each sub-stream Fm in the stream F. Alternatively, at the end of the coding step C3 above, the coder CO orders the substreams F1, F2, ..., Fm, ..., FL in the stream F, in the order expected by the decoder DO, which avoids the insertion in the stream F subflow indicators. Such an arrangement thus makes it possible to reduce the flow rate cost of the data stream F.

In the example shown in FIG. 4A or 4B, the said MB blocks have a square shape and all have the same size. Depending on the size of the image that is not necessarily a multiple of the size of the blocks, the last blocks on the left and the last blocks on the bottom may not be square. In an alternative embodiment, the blocks may be for example of rectangular size and / or not aligned with each other.

Such identification is performed by an EXDO flow extraction module as shown in FIG. 6A.

In the example shown in FIG. 4A or 4B, the predetermined number P is equal to 6 but only four subsets SE1, SE2, SE3, SE4 are shown in dashed line, for the sake of clarity of the figures.

With reference to FIG. 5A, the second decoding step D2 is the decoding of each of said subsets of blocks SE1, SE2, SE3 and SE4, the blocks of a subset considered being coded according to a sequential scanning order PS predetermined. In the example shown in FIG. 4A or 4B, the blocks of a current subset SEk (1 <k <P) are decoded one after the other, from left to right, as indicated by the arrow PS. . At the end of step D2, the subsets of decoded blocks SED1, SED2, SED3,..., SEDk,..., SEDP are obtained.

Such a decoding can be of sequential type and, consequently, be carried out using a single decoding unit.

However, in order to benefit from a multiplatform decoding architecture, the decoding of the subsets of blocks is of parallel type and is implemented by a number R of decoding units UDk (1 <k <R), with for example R = 4 as shown in Figure 6A. Such an arrangement thus allows a substantial acceleration of the decoding method. In a manner known per se, the decoder DO comprises a buffer MT which is adapted to contain the probabilities of appearance of symbols such as progressively updated as the decoding of a current block.

As shown in more detail in FIG. 6B, each of the UDk decoding units comprises:

An entropy decoding module of said current block by learning at least one computed symbol occurrence probability for at least one previously decoded block, denoted MDEk,

A decoding module predictive of a current block with respect to said previously decoded block, denoted MDPk.

The predictive decoding module SUDPk is able to perform a predictive decoding of the current block, according to conventional prediction techniques, such as for example in Intra and / or Inter mode.

The entropy decoding module MDEk is CABAC type, but modified according to the present invention as will be described later in the description.

Alternatively, the entropy decoding module MDEk could be a Huffman decoder known as such.

In the example shown in FIG. 4A or 4B, the first unit UD1 decodes the blocks of the first line SE1, from left to right. When it reaches the last block of the first line SE1, it passes to the first block of the (n + 1) ^th line, by the 5 ^th row, etc. The second unit UC2 decodes the blocks of the second line SE2, from left to right. When it reaches the last block of the second line SE2, it passes to the first block of the (n + 2) ^th line, by the 6 ^th row, etc. This path is repeated until the unit UD4, which decodes the blocks of the fourth line SE4, from left to right. When it reaches the last block of the first line, it passes the first block of the (n + 4) ^th line, here the 8 ^th line, and so on until the last block of the last substream identified is decoded.

Other types of course than the one just described above are of course possible. For example, each decoding unit might not handle nested lines, as explained above, but columns nested. It is also possible to browse the rows or columns in one direction or the other.

With reference to FIG. 5A, the third decoding step D3 is the reconstruction of a decoded picture ID from each decoded subset SED1, SED2,..., SEDk,..., SEDP obtained at the step decoding D2. More precisely, the decoded blocks of each decoded subset SED1, SED2,..., SEDk,..., SEDP are transmitted to an image reconstruction URI unit as represented in FIG. 6A. During this step D3, the URI unit writes the decoded blocks in a decoded image as these blocks become available.

During a fourth decoding step D4 shown in FIG. 5A, a fully decoded ID image is provided by the URI unit shown in FIG. 6A.

We will now describe, with reference to FIG. 5B, the different specific sub-steps of the invention, as implemented during the above-mentioned parallel decoding step D2, in a decoding unit UDk.

During a step D21, the decoding unit UDk selects as the current block the first block to be decoded from the current line SEk shown in FIG. 4A or 4B.

During a step D22, the decoding unit UDk tests whether the current block is the first block of the decoded picture, in this case the first block of the substream F1.

If this is the case, during a step D23, the entropy decoding module MDE or MDEk initializes its state variables. According to the example shown, it is an initialization of an interval representative of the probability of appearance of a symbol contained in the predetermined set of symbols.

Alternatively, if the entropy decoding used is LZW decoding, a symbol string translation table is initialized, so that it contains once and only once all the possible symbols. Since step D23 is identical to the aforementioned coding step C33, it will not be described further. If following the aforementioned step D22, the current block is not the first block of the decoded picture ID, it is proceeded, during a step D30 which will be described later in the following description, determining the availability of the previously decoded blocks needed.

During a step D24, the first current block MB1 of the first line SE1 shown in FIG. 4A or 4B is decoded. Such a step D24 comprises a plurality of substeps D241 to D246 which will be described below.

During a first substep D241, entropic decoding of the syntax elements related to the current block is performed. Such a step consists mainly of:

a) read the bits contained in the sub-stream associated with said first line SE1,

b) reconstruct the symbols from the read bits.

In the aforementioned variant according to which the decoding used is an LZW decoding, digital information corresponding to the code of the symbol in the current translation table is read, the symbol is reconstructed from the code read and an update of the translation table. is performed, according to a method known per se.

More precisely, the syntax elements related to the current block are decoded by the entrapment decoding module MDE1 CABAC as represented in FIG. 6B. The latter decodes the sub-stream of F1 bits of the compressed file to produce the syntax elements, and, at the same time, updates its probabilities so that, at the moment when the latter decodes a symbol, the probabilities of appearance of this symbol are identical to those obtained during the coding of the same symbol during the aforementioned entropy coding step C345.

During a following substep D242, predictive decoding of the current block MB1 is carried out by known intra and / or inter prediction techniques, during which the block MB1 is predicted with respect to at least one previously decoded block. . It goes without saying that other intra prediction modes as proposed in the H.264 standard are possible.

During this step, the predictive decoding is performed using the syntax elements decoded in the previous step and including in particular the type of prediction (inter or intra), and if appropriate, the intra prediction mode, the type of partitioning of a block or macroblock if the latter has been subdivided, the reference image index and the displacement vector used in the inter prediction mode.

Said aforementioned predictive decoding step makes it possible to construct a predicted MBp block! .

During a subsequent substep D243, a quantized residual block MBqi is constructed using previously decoded syntax elements.

During a subsequent substep D244, dequantization of the quantized residue block MBq is carried out according to a conventional dequantization operation which is the inverse operation of the quantization performed in the aforementioned step C344, to produce a dequantized block decoded MBDt-i.

During a subsequent substep D245, the inverse transformation of the dequantized block MBDti is carried out which is the reverse operation of the direct transformation carried out in step C343 above. A decoded residue block MBDr-ι is then obtained.

In the course of a subsequent substep D246, the decoded block MBDi is constructed by adding or predicting MBpi the decoded residue block MBDr-i. The decoded block MBD is thus made available for use by the decoding unit UD1 or any other decoding unit belonging to the predetermined number N of decoding units.

At the end of the aforementioned decoding step D246, the entropy decoding module MDE1 as represented in FIG. 6B contains all the probabilities such as progressively being updated as the first block is decoded. These probabilities correspond to the different possible syntax elements and the different decoding contexts associated with them. Following the aforementioned decoding step D24, it is tested, during a step D25, if the current block is the jth block of this same line, where j is a predetermined known value of the decoder DO which is at less than 1.

If this is the case, during a step D26, the set of probabilities calculated for the jth block is stored in the buffer memory MT of the decoder DO as represented in FIG. 6A and in FIG. 4A or 4B, FIG. size of said memory being adapted to store the calculated number of probabilities.

During a step D27, the unit UDk tests whether the current block that has just been decoded is the last block of the last sub-stream.

If this is the case, during a step D28, the decoding process is terminated.

If this is not the case, step D29 is used to select the next block MB to be decoded according to the order of travel represented by the arrow PS in FIG. 4A or 4B.

If during the aforementioned step D25, the current block is not the jth block of the line SEDk considered, it is proceeded to step D27 above.

During a step D30 which follows the aforementioned step D29, the availability of previously decoded blocks which are necessary for decoding the current block MB, is determined. Given that it is a parallel decoding of the blocks by different UDk decoding units, it is possible that these blocks have not been decoded by the decoding unit assigned to the decoding of these blocks. and so they are not yet available. Said determination step consists in checking whether a predetermined number N 'of blocks located on the previous line SEk-1, for example the two blocks situated respectively above and to the right of the current block, are available for the decoding of the current block, that is to say if they have already been decoded by decoding unit UDk-1 assigned to the decoding of the latter. Said determination step also consists in checking the availability of at least one block to the left of the current block to be decoded MB ,. However, given the PS travel order chosen in the embodiment shown in Figure 4A or 4B, the blocks are decoded one after the other on a line SEk considered. As a result, the block Decoded left is always available (except the first block of a line). In the example shown in FIG. 4A or 4B, it is the block located immediately to the left of the current block to be decoded. For this purpose, only the availability of the two blocks located respectively above and above the right of the current block is tested.

Since this test step is likely to slow down the decoding method, alternatively according to the invention, a clock CLK shown in FIG. 6A is adapted to synchronize the advance of the decoding of the blocks so as to guarantee the availability of the two blocks. located respectively above and above right of the current block, without it being necessary to check the availability of these two blocks. Thus, as represented in FIG. 4A or 4B, a decoding unit UDk always begins to decode the first block with an offset of a predetermined number N '(here N' = 2) of decoded blocks of the preceding line SEk-1 which are used for decoding the current block. From a software point of view, the implementation of such a clock makes it possible to significantly accelerate the processing time of the blocks of each subset SEk in the decoder DO.

During a step D31, it is tested if the current block is the first block of the SEk line considered.

If this is the case, during a step D32, only the probabilities of appearance of symbols calculated during the decoding of the jth block of the preceding line SEk-1 are read in the buffer memory MT.

According to a first variant represented in FIG. 4A, the jth block is the first block of the preceding line SEk-1 (j = 1). Such a reading consists in replacing the probabilities of the CABAC decoder by those present in the buffer memory MT. With regard to the first respective blocks of the second, third and fourth lines SE2, SE3 and SE4, this reading step is shown in FIG. 4A by the arrows represented in fine lines.

According to a second variant of the aforementioned step D32 which is illustrated in FIG. 4B, the jth block is the second block of the preceding line SEk-1 (j = 2). Such a reading consists in replacing the probabilities of the CABAC decoder by those present in the buffer memory MT. With regard to first respective blocks of the second, third and fourth lines SE2, SE3 and SE4, this reading step is shown in Figure 4B by the arrows shown in dashed fine lines.

Following step D32, the current block is decoded by iteration of the steps D24 to D28 described above.

If following the aforementioned step D31, the current block is not the first block of the line SEk considered, it is advantageously not proceeded to the reading of the probabilities from the previously decoded block which is on the same line SEk, that is to say the decoded block located immediately to the left of the current block, in the example shown. In fact, given the sequential reading path PS of the blocks located on the same line, as represented in FIG. 4A or 4B, the probabilities of appearance of symbols present in the CABAC decoder at the moment of the beginning of the decoding of the current block are exactly the ones that are present after decoding the previous block on that same line.

Consequently, during a step D33, the symbol occurrence probabilities are learned for the entropy decoding of said current block, which probabilities correspond only to those calculated for said previous block on the same block. line, as represented by the double arrows in solid lines in Figure 4A or 4B.

Following step D33, the current block is decoded by iterating the steps D24 to D28 described above.

It is then tested, during a step D34, if the current block is the last block of the line SEk considered.

If this is not the case, following step D34, step D29 for selecting the next block MB to be coded is again implemented.

If the current block is the last block of the line SEk considered, during a step D35, the decoding unit UDk performs a step identical to the aforementioned step D23, that is to say initializes again the an interval representative of the probability of appearance of a symbol contained in the predetermined set of symbols. Such a reset is shown in Figures 4A and 4B by a black dot at the beginning of each line SEk. Thus, the decoder DO is in a state initialized at each beginning of the line, which allows a great flexibility from the point of view of the choice of the level of decoding parallelism and an optimization of the processing time during decoding.

In the exemplary coding / decoding scheme shown in FIG. 7A, the coder CO comprises a single coding unit UC, as shown in FIG. 3A, while the decoder DO comprises six decoding units.

The encoding unit UC sequentially encodes the lines SE1, SE2, SE3, SE4, SE5 and SE6. In the example shown, the lines SE1 to SE4 are fully coded, the line SE5 is being coded and the line SE6 has not yet been coded. Given the sequentiality of the coding, the coding unit UC is adapted to deliver a stream F which contains the substreams F1, F2, F3, F4 ordered one after the other, in the coding order of the lines SE1, SE2, SE3 and SE4. For this purpose, the substreams F1, F2, F3 and F4 are symbolized with the same hatching as those which respectively symbolize the lines SE1, SE2, SE3, SE4 encoded. Thanks to the steps of emptying at the end of coding of said coded lines and the resetting of the interval of probabilities at the start of coding or decoding of the next line to be coded / decoded, the decoder DO, whenever it reads a sub-stream to decode it, is in an initialized state and can therefore optimally decode in parallel the four sub-streams F1, F2, F3, F4 with decoding units UD1, UD2, UD3 and UD4 which can by example be installed on four different platforms.

In the exemplary coding / decoding scheme shown in FIG. 7B, the coder CO comprises two coding units UC1 and UC2, as shown in FIG. 3C, while the decoder DO comprises six decoding units.

The encoding unit UC1 sequentially codes the odd-rank lines SE1, SE3 and SE5, while the encoding unit UC2 sequentially encodes the even-numbered lines SE2, SE4 and SE6. For this purpose, lines SE1, SE3 and SE5 have a white background, while lines SE2, SE4 and SE6 have a dotted background. In the example shown, the lines SE1 to SE4 are fully encoded, the SE5 line is being coded and the SE6 line has not yet been coded. Given the fact that the coding performed is of parallel type of level 2, the coding unit UC1 is adapted to deliver a sub-flux F _{2n +} i broken down into two parts F1 and F3 obtained following the coding respectively of the lines SE1 and SE3 , while the coding unit UC2 is adapted to deliver a sub-flux F _2n decomposed into two parts F2 and F4 obtained following the coding respectively lines SE2 and SE4. The encoder CO is therefore adapted to transmit to the decoder DO a stream F which contains the juxtaposition of the two substreams F _{2n +} 1 and F _2n and therefore a scheduling of the substreams F1, F3, F2, F4 which differs from that shown in FIG. 7A. For this purpose, the substreams F1, F2, F3 and F4 are symbolized with the same hatching as those which respectively symbolize the lines SE1, SE2, SE3, SE4 coded, the substreams F1 and F3 having a white background (coding lines of odd rank) and the substreams F2 and F4 having a dotted background (coding of lines of even rank).

With respect to the advantages mentioned in connection with FIG. 7A, such a coding / decoding scheme also has the advantage of being able to have a decoder whose level of decoding parallelism is completely independent of the level of parallelism of the coding. which makes it possible to further optimize the operation of an encoder / decoder.

Claims

1. A method of coding at least one image comprising the steps of:

- cutting (C1) the image into a plurality of blocks (MB) able to contain symbols belonging to a predetermined set of symbols,

grouping (C2) blocks into a predetermined number (P) of block subsets (SE1, SE2, ..., SEk, ..., SEP),

coding (C3), by means of an entropy coding module, of each of said subsets of blocks, by association of digital information with the symbols of each block of a subset considered, said encoding step comprising, for the first block of the image, a substep (C33) for initializing state variables of said entropy coding module,

generating (C4) at least one data sub-stream (F1) representative of at least one (SE1) of said coded block subsets, said coding method being characterized in that:

in the case where the current block is the first block to be encoded of a subset considered, determining (C42) symbol appearance probabilities for said first current block, said probabilities being those which have been determined for a block predetermined coded and decoded at least one other subset,

in the case where the current block is the last coded block of the subset considered:

Writing (C45), in the representative sub-stream of said at least one subset considered, of all the digital information that has been associated with the symbols during the coding of the blocks of said subset considered,

Implementation (C46) of said initialization substep.

2. Coding method according to claim 1, in which the subsets of blocks (SE1, SE2, ..., SEk, ..., SEP) are encoded sequentially or in parallel.

3. Method according to claim 2, in which when at least two subsets of blocks (SE1, SE3) are coded parallel to at least one other subset of blocks (SE2), said at least two subsets. coded blocks are contained in the same data sub-stream

Coding method according to any one of claims 1 to 3, wherein when said subsets of coded blocks are to be decoded in parallel in a predetermined order, the sub-data streams delivered after coding respectively of each of said subsets of blocks are previously ordered according to said predetermined order before being transmitted for decoding.

5. Encoding device (CO) of at least one image comprising:

means (PCO) for cutting the image into a plurality of blocks (MB) capable of containing symbols belonging to a predetermined set of symbols,

means (GRCO) for grouping said blocks into a predetermined number (P) of block subsets,

means (UC1, UC2,..., UCk,..., UCP) encoding each of said block subsets, said coding means comprising an entropy coding module (MCE; MCEk) adapted to associating digital information with the symbols of each block of a subset considered, said coding means comprising, for the first block of the image, sub-means for initializing state variables of said entropy coding module,

means (MGSF; MGSFk) for generating at least one data sub-stream (F1) representative of at least one of said coded block subsets, said coding device being characterized in that it comprises:

means for determining symbol appearance probabilities for a current block which, in the case where the current block is the first block to be coded for a considered subset, determine the symbol appearance probabilities for said first block block as those which have been determined for a predetermined coded and decoded block of at least one other subset,

writing means which, in the case where the current block is the last coded block of said subset considered, are activated to write, in the representative sub-stream of said subset considered, all the digital information which has have been associated with the symbols during the coding of the blocks of said subset considered,

said initialization sub-means being further enabled to reset the state variables of said entropy coding module.

A computer program comprising instructions for implementing the encoding method according to any one of claims 1 to 4, when executed on a computer.

A computer-readable recording medium on which a computer program is recorded including instructions for executing the steps of the encoding method according to any one of claims 1 to 4, when said program is executed by a computer.

8. A method for decoding a stream (F) representative of at least one coded picture, comprising the steps of:

identification (D1) in said stream of a predetermined number of data substreams (F1, F2, ..., Fm, ..., FL) respectively corresponding to at least one subset of blocks to be decoded, said blocks being able to contain symbols belonging to a predetermined set of symbols,

decoding (D2) said identified subsets of blocks by means of an entropy decoding module, by reading, in at least one of said identified substreams, digital information associated with the symbols each block of the subset corresponding to said at least one identified substream, said decoding step comprising, for the first block to be decoded from the image, a substep (D23) for initializing state variables of said entropy decoding module,

said decoding method being characterized in that:

in the case where the current block is the first block to be decoded from a subset considered, determination (D32) of symbol appearance probabilities for the first block of said subset considered, said probabilities being those which have been determined for a predetermined decoded block of at least one other subset,

in the case where the current block is the last decoded block of the subset considered, implementation (D35) of said initialization substep.

9. decoding method according to claim 8, wherein the subsets of blocks are decoded sequentially or in parallel.

The method of claim 9, wherein when at least two subsets of blocks are decoded in parallel with at least one other subset of blocks, one of the identified data substreams is representative of the at least two subsets of blocks. - blocks sets.

1 1. A decoding method according to claim 9 or claim 10, wherein when said coded block subsets are to be decoded in parallel in a predetermined order, the data sub-streams respectively corresponding to said coded block subsets. are ordered beforehand according to said predetermined order in said flow to be decoded.

12. Device (DO) for decoding a stream (F) representative of at least one coded picture, comprising:

means (EXDO) for identifying in said stream a predetermined number of corresponding sub-data streams (F1, F2,..., Fm,..., FL) respectively to at least a subset of blocks to be decoded, said blocks being able to contain symbols belonging to a predetermined set of symbols,

means (UD, UD1, UD2,..., UDk,... UDP) for decoding said identified subsets of blocks, said decoding means comprising an entropy decoding module (MDE; MDE1, MDE2, .. ., MDEk, ..., MDEP) able to read, in at least one of said identified substreams, digital information associated with the symbols of each block of the subset corresponding to said at least one identified substream, said means of decoding comprising, for the first block to be decoded from the image, sub-means for initializing state variables of said entropy decoding module,

said decoding device being characterized in that it comprises means for determining symbol occurrence probabilities for a current block which, in the case where the current block is the first block to be decoded of a considered subset, determine the symbol occurrence probabilities for said first block as those determined for a decoded predetermined block of at least one other subset,

and in that in the case where the current block is the last decoded block of the subset considered, said initialization sub-means are activated to reset the state variables of said entropy decoding module.

Computer program comprising instructions for implementing the decoding method according to any one of claims 8 to 11 when executed on a computer.

14. Computer-readable recording medium on which is recorded a computer program comprising instructions for executing the steps of the decoding method according to any one of claims 8 to 11, when said program is executed by a computer.