RU2782400C2 - Method of encoding and decoding images, device for encoding and decoding and corresponding software - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: invention relates to image encoding and decoding. Disclosed is a method of decoding a portion (Bi) of an image divided into portions, comprising: receiving a set of coefficients encoded by a context adaptive binary arithmetic coding (CABAC) type encoder; decoding a set of coefficients representing a block of residual data of the portion, wherein the first non-zero coefficient does not contain any sign value; applying a function to a set of coefficients to generate sign data, wherein applying the function to the set of coefficients comprises: calculating the sum of coefficients, which are not equal to zero, from said set of coefficients; calculating parity of said sum of coefficients. Decoding method is characterized by that a sign is assigned to a first non-zero coefficient based on parity, wherein the even value of said sum of coefficients means that the sign of the first non-zero coefficient is positive, and the odd value of said sum of coefficients means that the sign of the first non-zero coefficient is negative.

EFFECT: high accuracy of encoding and decoding images.

4 cl, 4 dwg

Description

The technical field to which the invention belongs

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

The invention is thus particularly intended for video coding as embodied in current (MPEG, H.264, etc.) or future video encoders (ITU-T/VCEG (H.265) or ISO/MPEG (HEVC) ).

State of the art

Modern video encoders (MPEG, H.264, etc.) use a representation in the form of video sequence blocks. Images are divided into macroblocks, each macroblock is itself divided into blocks, and each block or macroblock is encoded using intra-image or inter-image prediction. Thus, some pictures are encoded using spatial prediction (intra-picture prediction), while other pictures are encoded using temporal prediction (inter-picture prediction) on one or more reference coded-decoded pictures using motion compensation, as is known to those skilled in the art. this field of technology.

For each block, a residual block, also called a prediction residual, corresponding to the original block reduced during the prediction, is encoded. The residual blocks are transformed using a discrete cosine transform (DCT) type transform, then quantized using, for example, scalar type quantization. At the end of the quantization step, coefficients are obtained, some of which are positive and others are negative. They are then run through, usually in a zigzag reading order (as in the JPEG standard), thus making it possible to use a significant number of zero coefficients at high frequencies. At the end of the above run, a one-dimensional list of coefficients is obtained, which can be called the "quantized residual". The coefficients from this list are then encoded using entropy coding.

Entropy coding (for example, arithmetic coding or Huffman type coding) is performed as follows:

- the information element is subjected to entropy encoding to indicate the position of the last non-zero coefficient in the list,

- for each coefficient located before the last non-zero coefficient, the information element is entropy encoded to indicate whether this coefficient is zero or not,

- for each non-zero coefficient specified earlier, an information element is entropy encoded to indicate whether this coefficient is equal to one or not,

- for each coefficient that is not equal to zero and not equal to one and which was located before the last non-zero coefficient, the amplitude information element (the absolute value of the coefficient whose value has been halved) is entropy encoded,

- for each non-zero coefficient, the sign assigned to it is encoded as "0" (for the sign "+") or as "1" (for the sign "-").

According to H.264 technology, for example, when a macroblock is divided into blocks, a data signal corresponding to each block is transmitted to a decoder. This signal contains:

- quantized residuals contained in the list mentioned above,

- information elements representing the encoding mode used, in particular:

• prediction mode (intra-frame prediction, inter-frame prediction, default prediction, which obtains a prediction for which no information is sent to the decoder (called "skip");

• information specifying the type of prediction (orientation, reference image, etc.);

• type of division into sections;

• transformation type, eg 4x4 DCT, 8x8 DCT, etc.;

• traffic information, if necessary;

• etc.

Decoding is performed picture by picture, and for each picture macroblock by macroblock. For each section of the macroblock, the corresponding elements of the stream are read. Dequantization and inverse transformation of the block coefficients are performed to obtain a decoded prediction residual. The segment prediction is then calculated, and the segment is reconstructed by adding the prediction to the decoded prediction residual.

Encoding within a frame, or between frames, as a result of competition with each other, as embodied in the H.264 standard, is thus based on placing various coding information elements, such as those mentioned above, in such a way that they compete to select the best mode, then is, a mode that optimizes the coding of the area under consideration, in accordance with a given performance criterion, for example, the cost of random distortion, as is well known to a person skilled in the art.

Information elements representing the selected coding mode are contained in the data signal transmitted from the encoder to the decoder. The decoder is thus configured to identify the coding mode selected in the encoder and then apply the prediction that corresponds to that mode.

In "Data Hiding of Motion Information in Chroma and Luma Samples for Video Compression", J.-M. Thiesse, J. Jung and M. Antonini, International workshop on multimedia signal processing, 2011 presents a data hiding method that is implemented during video data compression.

More specifically, it is proposed to avoid including in a signal to be transmitted to a decoder at least one contention index generated from a plurality of contention indices to be transmitted. Such an index is, for example, an MVComp index that represents an information element for identifying a motion vector predictor used for an inter prediction block. Such an index, which may take on the value of 0 or 1, is not included directly in the encoded cell signal, but is transmitted by the parity of the sum of the coefficients of the quantized residual. A relationship is formed between the parity of the quantized residual and the MVComp index. As an example, an even quantized residual value is associated with an MVComp index value of 0, while an odd quantized residual value is associated with an MVComp index with a value of 1. Two cases may arise. In the first case, if the parity of the quantized residue already matches the parity of the MVComp index to be transmitted, the quantized residue is encoded in the usual way. In the second case, if the parity of the quantized residual differs from that of the MVComp index to be transmitted, the quantized residual is modified so that its parity is the same as that of the MVComp index. remainder by an odd value (eg, +1, -1, +3, -3, +5, -5, etc.) and maintain only the modification that optimizes the given criterion, in this case, the previously mentioned cost of random distortion.

In the decoder, the MVComp index is not read in the signal. The decoder is content with just the usual definition of the remainder. If the value of such a remainder is even, the MVComp index is set to 0. If the value of this remainder is odd, the MVComp index is set to 1.

In accordance with the technique that has been presented above, the coefficients that are modified are not always chosen optimally, such that the modification applied results in disturbances in the signal transmitted to the decoder. Such violations inevitably have a negative impact on the efficiency of video data compression.

Disclosure of invention

One of the objectives of the invention is to overcome the disadvantages of the prior art mentioned above.

Thus, the purpose of the present invention relates to a method for encoding at least one image divided into sections, such a method implements the following steps:

- perform the prediction of the data elements of the current section as a function of at least one reference section already encoded and then decoded, outputting the predicted section;

- determining a set of residual data elements by comparing data elements related to the current site and the predicted site, and the residual data elements are respectively associated with different digital information elements for which it is required to perform entropy coding,

- generate a signal containing encoded information elements. The method according to the invention is characterized by the fact that it implements, before

signal generation step, the following steps:

- determine from a certain set of residual data elements a subset containing residual data elements suitable for modification,

- calculate the value of the function representing the elements of the residual data of the specified subset,

- comparing the calculated value with the value of at least one of the digital information elements,

- depending on the result of the comparison, at least one of the residual data elements in the subset is modified or not modified,

- in case of modification, entropy encoding of said at least one element of the modified residual data is performed.

This sequence allows the data hiding technology to be applied in a reduced set of residual data elements, and in this set the residual data elements are suitable for their modification.

In accordance with the invention, it should be understood that the expression "residual data elements" means data elements for which the application of the modification does not lead to desynchronization between the encoder and decoder.

Thus, according to the invention, the residual data elements to be modified are much more reliably selected than in the previously cited prior art, thus allowing a better image reconstruction quality to be obtained in the decoder.

In addition, the possibility of modifying a reduced number of residual data items enables faster coding.

In a specific embodiment, the steps that follow the step of determining a subset of residual data items are only implemented if a predetermined criterion is satisfied, depending on the residual data items suitable for modification.

Such a sequence further enables the encoder to rationally determine whether or not to apply a hiding technique to the data. Such a determination step has the advantage of applying it only to a limited set of residual data elements suitable for their modification. Thus, it is ensured that the data hiding technology is applied much more appropriately than in the previously cited prior art, in particular for the best selected number of residual data items, for which it is precisely determined that after modifying these data items, the violation of the signal generated as a result of such a modification, will not adversely affect the quality of the image reconstruction in the decoder.

In another particular embodiment, the predetermined determination criterion depends on the result of the comparison between the number of residual data elements suitable for their modification and the predetermined number.

Such a sequence improves the compression performance of the arithmetic encoder while effectively reducing signaling costs. In particular, such a sequence makes it possible to accurately detect the number of residual data elements from which it would be advisable to apply the data hiding technology so that the latter does not lead to high levels of disturbances in the signal intended for transmission to the decoder.

In yet another particular embodiment, if a plurality of digital information items are considered during the comparison step, the latter results in a comparison of a calculated function value representing the residual data items of a particular subset with a function value representing the plurality of digital information items.

Such a sequence optimizes the performance of the arithmetic encoder while optimizing the reduction of signaling costs, since it provides for the hiding of several elements of digital information in the signal intended for transmission to the decoder.

In another particular embodiment, at least one element of digital information corresponds to the sign of the residual data element.

The sign is a particularly valuable piece of information that needs to be hidden due to the fact that the probability of a positive or negative sign is equally probable. Also, given that the sign is necessarily encoded in one bit, it thus becomes possible, by hiding this information, to save one bit in the signal for transmission to the decoder, thus significantly reducing the signaling cost. The reduction in such cost will be all the greater when it is possible, according to the invention, to hide a plurality of characters, and therefore a plurality of bits.

Accordingly, the invention also relates to a device for encoding at least one image divided into sections, such a device comprising:

means for predicting data elements of the current section as a function of at least one reference section already encoded with subsequent decoding to provide the predicted section,

- means for determining a set of residual data elements suitable for comparing data elements relating to the current site and the predicted site, the residual data elements are associated, respectively, with different digital information elements for which entropy encoding is required,

- a means of generating a signal containing elements of encoded information.

Such an encoder is characterized in that it comprises, before the generating means, a processing means which is configured to:

- to determine, from a certain set of residual data elements, a subset containing residual data elements suitable for their modification,

- calculate the value of the function representing the elements of the residual data of a certain subset,

- compare the calculated value with the value of at least one of the digital information elements,

- modify or not modify said at least one of the residual data elements of a certain subset, depending on the result of the modification,

and means for entropy encoding at least one element of the modified residual data in case of modification by the processing means.

Accordingly, the invention also relates to a method for decoding a data signal representing at least one image divided into regions that has been previously encoded, comprising the step of deriving, by entropy decoding of the signal data, digital data elements associated with residual data elements that are at least least to one previously encoded section.

This decoding method is characterized by the fact that it contains the following

stages:

- determine, from the elements of the residual data, a subset containing the elements of the residual data, made with the possibility of modifying them during the previous encoding,

- calculate the value of the function representing the elements of the residual data of the mentioned specific subset,

- from the calculated value determine the value of at least one element of digital information, which differs from those obtained by entropy decoding.

In a particular embodiment, the steps that follow the step of determining a subset of residual data items are only implemented if a predetermined criterion is satisfied, depending on the residual data items to be modified.

In another particular embodiment, the predetermined determination criterion depends on the result of the comparison between the number of residual data items that could be modified and the predetermined number.

In yet another particular embodiment, a plurality of values associated respectively with a plurality of digital data elements that differ from those obtained as a result of entropy decoding are obtained from said calculated value.

In yet another particular embodiment, said at least one digital information element corresponds to the sign of the residual data element.

Accordingly, the invention also relates to a device for decoding a data signal representing at least one division of an image into areas that have been previously encoded, comprising means for obtaining, by entropy decoding of signal data elements, digital information elements associated with residual data elements, in relation to at least at least one previously coded section.

Such a decoding device is characterized in that it comprises a processing means which is configured to:

determine, from said residual data elements, a subset containing residual data elements that may have been modified during previous encoding,

- calculate the value of the function representing the elements of the residual data of a certain subset,

- from the calculated value to obtain the value of at least one element of digital information, which differs from those obtained as a result of entropy decoding.

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

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

Yet another aspect of the invention is a computer-readable recording medium including computer program instructions as mentioned above.

The recording medium may be any object or device capable of storing the program. For example, such a medium may include a storage medium such as a ROM, such as a CD-ROM or ROM in microelectronic circuits, or a magnetic recording medium, such as a floppy disk or hard disk.

On the other hand, such a recording medium may be a transmission medium, such as an electrical or optical signal, which may be transmitted through an electrical or optical cable, through a radio channel, or through other means. The program according to the invention can in particular be downloaded via a network such as the Internet.

Alternatively, such a recording medium may be an integrated circuit in which the program is embedded, the circuit is capable of executing the method in question or can be used in executing the latter.

The encoding device, decoding method, decoding device and computer programs mentioned above exhibit at least the same advantages as those provided by the encoding method according to the present invention.

Brief description of the drawings

Other features and advantages will become apparent on reading the two preferred embodiments described with reference to the drawings, in which:

in fig. 1 shows the main steps of the encoding method according to the invention,

in fig. 2 shows an embodiment of a decoding device according to the invention,

in fig. 3 shows the main steps of the decoding method according to the invention,

in fig. 4 shows an embodiment of a decoding device according to the invention.

Detailed description of the coding part

An embodiment of the invention will be described below in which an encoding method according to the invention is used to encode a sequence of images according to a bit stream close to that obtained by encoding according to the H.264/MPEG-4 AVC standard. In this embodiment, the coding method according to the invention is, for example, embodied in software or in hardware form by modifications to an encoder originally conforming to the H.264/MPEG-4 AVC standard. The encoding method according to the invention is in the form of an algorithm including steps C1 to C40 shown in FIG. one.

According to an embodiment of the invention, the encoding method according to the invention is embodied in an encoder or CO encoder, an embodiment of which is shown in FIG.

According to the invention, before the actual encoding step, the image sequence IE to be encoded in a predetermined order is divided into a plurality Ζ of sections Β ₁ , B ₂ , ..., B _i , ..., B _z , as shown in FIG. 2.

It should be noted that in the sense of the invention, the term "plot" means a coding unit. This latter terminology is in particular used in the HEVC/H.265 standard, a draft of which, for example, is provided in a document available at the following Internet address:

<http://phenix.int-evry.fr/jct/doc_end_user/current_document.php?id=3286>

In particular, such groups of coding units together set up rectangular or square shaped pixels, also called blocks, macroblocks, or pixel sets, which have other geometric shapes.

In the example shown in FIG. 2, said areas are blocks which are square in shape and all of which are of the same size. Depending on the size of the image, which is not necessarily a multiple of the block size, the last blocks on the left and the last blocks on the bottom need not be square. In an alternative embodiment, the blocks may, for example, be rectangular and/or may not be aligned with each other.

Each block or macroblock can furthermore be itself subdivided into subblocks, which themselves can then be subdivided.

This separation is performed using the separation module PCO shown in FIG. 2, which uses, for example, a partitioning algorithm well known per se.

After said separation step, each of the current sections B is encoded_i (where i is an integer such as 1 ≤ i ≤ Ζ) of said image IE.

In the example shown in FIG. 2, such coding is applied sequentially for each of the blocks Β ₁ -B _Z of the current picture IE. The blocks are encoded, for example, in accordance with a scan, such as a raster scan, which is known to a person skilled in the art.

The coding according to the invention is implemented in the MC_CO encoding software module of the CO encoder, as shown in FIG. 2.

During step C1 shown in FIG. 1, the MC_CO encoding module of FIG. 2 selects, as the current block B ₁ , the first block B ₁ to encode the current picture IE. As shown in FIG. 2, it is the first block to the left of the IE image.

During step C2 shown in FIG. 1, prediction encoding of the current block B ₁ is performed using known intra-frame and/or inter-frame prediction techniques, during which prediction is performed using block prediction encoding Β ₁ on at least one previously encoded and decoded block. Such prediction is performed using the prediction software module PRED_CO as shown in FIG. 2.

Needless to say, other modes of intra-prediction proposed in the H.264 standard are possible.

For the current block Β ₁ , it is also possible to perform inter-frame prediction encoding, during which the current block is predicted with respect to a block from a previously encoded and decoded picture. Other types of prediction may also be provided. Among the prediction options possible for the current block, the optimal prediction is selected in accordance with random distortion criteria that are known to a person skilled in the art.

Said predictive encoding step enables predictive encoding of a block Bp ₁ that is an approximation of the current block B ₁ . Information elements related to such predictive coding will be included in the signal for transmission to the decoder. Such information elements contain, in particular, the type of prediction (inter-frame or intra-frame) and, if necessary, the intra-frame prediction mode, the types of division of the block or macroblocks, if the latter were divided, the reference picture index and the motion vector that was used in the inter-frame prediction mode. . Such information elements are compressed by the CO encoder.

During the next step C3 shown in FIG. 1, the prediction module PRED_CO compares the data items related to the current block Β ₁ with the data items of the prediction block Bp ₁ . More specifically, during this step, usually the prediction block Βp ₁ is subtracted from the current block B ₁ to obtain a residual block Br ₁ .

During the next step C4 shown in FIG. 1, the residual block Br ₁ is transformed according to a conventional direct transformation operation, such as, for example, the discrete cosine transform type DCT, to obtain the transformed block Bt ₁ . Such an operation is performed by the conversion software module MT_CO as shown in FIG. 2.

During the next step C5 shown in FIG. 1, the transformed block Btj is quantized according to a conventional quantization operation such as, for example, scalar quantization. Then a block Bq _{1 of} quantized coefficients is obtained. Such a step is performed using the quantization software module MQ_CO as shown in FIG. 2.

During the next step Sb shown in FIG. In the example shown, such a sweep is a conventional zigzag sweep. This step is performed using the readout program module ML_CO as shown in FIG. 2. At the end of step C6, a one-dimensional list Ε ₁ =(ε1, ε2, ..., εL) of coefficients is obtained, commonly known as the "quantized remainder", where L is an integer greater than or equal to 1. Each of the coefficients in the list Ε ₁ is interconnected with various elements of digital information for which entropy coding is supposed to be performed. Such digital information items are described below by way of example.

Assume that in the example shown, L=16, and that the list E ₁ contains the following sixteen coefficients: Ε ₁ =(0, +9, -7, 0, 0, +1, 0, -1, +2, 0, 0, +1, 0, 0, 0, 0).

In this case:

- for each coefficient before the last non-zero coefficient in the list Ε ₁ , an element of digital information such as a bit must be entropy encoded to indicate whether this coefficient is zero or not: if this coefficient is zero, for example, a bit with the value 0, while if the coefficient is not equal to zero, the bit with the value 1 will be encoded;

- for each non-zero coefficient +9, -7, +1, -1, +2, +1, an element of digital information such as a bit must be entropy encoded to indicate whether the absolute value of the coefficient is equal to one or not: if it is equal to 1, then, for example, a bit value with a value of 1 will be encoded, while if it is not equal to 1, a bit with a value of 0 will be encoded;

- for each non-zero coefficient and for which the absolute value is not equal to one, and which is located before the last non-zero coefficient, such as coefficients with values +9, -7, +2, the amplitude information element (the absolute value of the coefficient from which the value of two was subtracted ) is subjected to entropy encoding;

- for each non-zero coefficient, the sign assigned to it is encoded using a digital information element such as a bit, for example set to '0' (for a "+" sign) or set to '1" (for a "-" sign).

Below, with reference to FIG. 1, specific encoding steps will be described in accordance with the invention.

In accordance with the invention, it is determined that entropy encoding of at least one of the above-mentioned information elements should be excluded. For the reasons explained above in the description, in the preferred embodiment, it is determined that entropy encoding should not be performed for at least one sign of one of the mentioned coefficients in the list E ₁ .

As an alternative example, in particular, one could determine the need for entropy encoding at least the least significant bit of the binary representation of the amplitude of the first non-zero coefficient in said list E ₁ .

To this end, during step C7 shown in FIG. 1 select the number of characters to be hidden when performing a later entropy encoding step. Such a step is performed using the processing software module MTR_CO as shown in FIG. 2.

In a preferred embodiment, the number of characters to be hidden is one or zero. In addition, according to the said preferred embodiment, it is intended to hide the sign of the first non-zero coefficient, which is intended to be hidden. Therefore, in the presented example, the sign of the coefficient ε2= +9 is hidden.

In an alternative embodiment, the number of characters to be hidden is either zero, one, two, three or more.

According to a preferred embodiment, in step C7, during the first sub-step C71 shown in FIG. 1, the sublist SE ₁ containing the coefficients suitable for their modification, ε'1, ε'2, ..., ε'M, where Μ<L, determined from the mentioned list E ₁ . Such coefficients will be referred to below as modified coefficients.

In accordance with the invention, a coefficient is modifiable if modification of its quantized value does not result in a desynchronization in the decoder when the modified coefficient is processed by the decoder. Thus, the processing module MTR_CO is initially configured not to perform the modification:

- a coefficient of zero or coefficients placed before the first list of non-zero coefficients in such a way that the decoder does not affect the value of the hidden sign in this or in these zero coefficients,

- and for reasons of computational complexity, zero coefficient or coefficients located after the last non-zero coefficient.

In the example shown, at the end of sub-step C71, a sublist of SE ₁ is obtained such that SE ₁ =(9, -7, 0, 0, 1, 0, -1, 2, 0, 0, 1). Therefore, eleven modified coefficients are obtained.

During the next sub-step C72 shown in FIG. 1, the processing module MTR_CO proceeds to compare the number of modified coefficients with a predetermined threshold value TSIG. In the preferred embodiment, TSIG is 4.

If the number of modified coefficients is less than the threshold value TSIG, then during step C20 of FIG. 1, conventional entropy coding of the coefficients in the list E ₁ is performed, such as is performed, for example, in the CABAC encoder denoted by reference CE_CO in FIG. 2. For this purpose, the sign of each non-zero coefficient in the list E ₁ is entropy encoded.

If the number of modified coefficients is greater than the TSIG threshold, then during step C8 of FIG. 1, the processing module MTR_CO calculates a function value f that represents the coefficients in sublist SE ₁ .

In the preferred embodiment, in which it is intended to hide only one character in the signal intended for the decoder, the function f is the parity of the sum of the coefficients in sublist SE ₁ .

During step C9 shown in FIG. 1, the processing unit MTR_CO checks whether the parity of the sign value to be hidden matches the parity of the sum of the coefficients in the sublist SE ₁ according to a predetermined condition set in the CO encoder.

In the proposed example, the convention is such that a positive sign is associated with a bit with a value of zero, while a negative sign is associated with a bit with a value of one.

If, according to the condition adopted in the CO encoder according to the invention, the sign is positive, which corresponds to the coding bit value of zero, and if the sum of the coefficients in sublist SE ₁ is even, then step C20 is performed to entropy-code the coefficients in said above the list Ε ₁ , except for the sign of the coefficient ε2.

If, still, according to the convention adopted in the encoder CO, in accordance with the invention, the sign is negative, which corresponds to the value of the encoding bit equal to one, and if the sum of the coefficients in sublist SE ₁ is odd, then also perform step C20 for entropy the coding of the coefficients in the list Ε ₁ mentioned above, except for the sign of the coefficient ε2.

If, in accordance with the convention adopted in the encoder CO, in accordance with the invention, the sign is positive, which corresponds to the value of the encoding bit equal to zero, and if the sum of the coefficients in sublist SE ₁ is odd, then during step C10 shown in FIG. .1 modify at least one modifiable coefficient in sublist SE ₁ .

If, still in accordance with the convention adopted in the encoder CO, in accordance with the invention, the sign is negative, which corresponds to the value of the encoding bit equal to one, and if the sum of the coefficients in sublist SE ₁ is even, then also at step C10 modify at least at least one modifiable factor in sublist SE ₁ .

Such a modification operation is performed by the processing unit MTR_CO of FIG. 2.

In an exemplary embodiment in which SE ₁ =(+9, -7, 0, 0,+1, 0, -1, +2, 0, 0, +1), the total sum f of the coefficients is 5, and therefore , it is odd. In order for the decoder to be able to reconstruct the positive sign assigned to the first non-zero coefficient ε2=+9 without the CO encoder to pass this coefficient to the decoder, the parity of the sum must become even. Therefore, the processing module MTR_CO tests, during said step C10, various modifications of the coefficients in sublist SE ₁ , all with the aim of changing the parity of the sum of the coefficients. In a preferred embodiment, +1 or -1 is added to each modified coefficient, and the modification is selected from among these resulting modifications,

In a preferred embodiment, this selection results in an optimal prediction according to a performance criterion, which is, for example, a random distortion criterion, as is well known to one skilled in the art. Such a criterion is expressed by equation (1) below:

(1) J=D+λR

where D is the distortion between the original macroblock and the reconstructed macroblock, R is the cost of encoding in bits of encoding information elements, and λ is the Lagrange multiplier, the value of which may be fixed before encoding.

In the proposed example, the modification that leads to optimal prediction, in accordance with the random distortion criterion mentioned above, is the addition of the value 1 to the second coefficient -7 in the sublist SE ₁ .

Therefore, at the end of step C10, the modified sublist receives the modified sublist SEm ₁ =(+9, -6, 0, 0, +1, 0, -1, +2, 0, 0, +1).

It should be noted that during this phase, certain modifications are prohibited. Thus, if the first non-zero coefficient ε2 had a value of +1, it would be impossible to add -1 to it, since it would then be equal to zero, and thus lose its characteristic of the first non-zero coefficient in the list E ₁ . The decoder must then associate the decoded sign (by calculating the parity of the sum of the coefficients) with another coefficient, and this would cause a decoding error.

During the ethane C11 shown in FIG. 1, the processing module MTR_CO performs the corresponding modification in the list Ε ₁ . Then the following modified list Em ₁ =(0, +9, -6, 0, 0, +1, 0, -1, -+2, 0, 0, +1, 0, 0, 0, 0) is obtained.

Then, in step C20, to entropy-code the coefficients in the above list, Em ₁ is performed except for the sign of the coefficient ε2, which is the "+" sign of the coefficient 9 in the present example, and this sign is hidden in the evenness of the sum of the coefficients.

It should be noted that the amplitude set of the coefficients in the list Ε ₁ , or in the modified list Εm ₁ , is encoded before the character set, except for the sign of the first non-zero coefficient ε2, which is not encoded as explained above.

During the next step C30 shown in FIG. 1, the MC_CO encoding module of FIG. 2 tests whether the current encoding block is the last image block of the IE.

If the current block is the last block of the 1E image, then during step C40. shown in FIG. 1, the encoding method ends.

If this is not the case, the next block Bi is selected, which is then encoded according to the order of the previously mentioned bitmap by repeating steps C1 to C20, for 1 ≤ i ≤ Z.

After entropy coding of all B ₁ -B _Z blocks, a signal F is constructed representing, in binary form, said encoded blocks.

The construction of the binary signal F is embodied in the stream construction software module CF as shown in FIG. 2.

Stream F is then transmitted via a data network (not shown) to the remote terminal. The latter includes a decoder, which will be described in more detail later in this description.

Next, referring generally to FIG. 1, another embodiment of the invention will be described.

This other embodiment differs from the previous one only in the number of coefficients to be hidden, which is either 0 or N, where N is an integer such as N≥2.

To this end, the comparison sub-step C72 mentioned above is replaced by the sub-step C72a represented by the dotted line in FIG. 1, during which the modified coefficient set is compared with several predetermined thresholds 0<TSIG_1<TSIG_2<TSIG_3..., such that if the modified coefficient set is between TSIG_N and TSIG_N+1, N characters are supposed to be hidden.

If the set of coefficients to be modified is less than the first threshold value TSIG_1, then during the above step C20, the normal entropy encoding of the coefficients in the list E ₁ is performed. For this purpose, the sign of each non-zero coefficient in the list E ₁ is subjected to entropy coding.

If the modified coefficient set is between the thresholds TSIG_N and TSIG_N+1, then during step C8 shown in FIG. 1, the processing module MTR_CO calculates a function value f that represents the coefficients in sublist E ₁ .

In this other embodiment, since the decision in the encoder is to hide N characters, the function f is the remainder of modulo 2 ^N of the sum of the coefficients in sublist SE ₁ . It is assumed, in the proposed example, that N=2, the two signs to be hidden are the first two signs of the first two non-zero coefficients, respectively, that is, ε2 and ε3.

During the next step C9 shown in FIG. 1, the processing module MTR_CO checks whether the pattern of N characters, that is, 2 ^N possible patterns, corresponds to the remainder value of module 2 ^N of the sum of the coefficients in sublist SE ₁ .

In the proposed example, where N=2, there are 2 ² =4 different sign configurations.

These four configurations correspond to a convention in the CO encoder, which convention is defined, for example, as follows:

- the remainder equal to zero corresponds to two consecutive positive signs: +, +;

- the remainder equal to one corresponds to a consecutive positive sign and a negative sign: +, -;

- the remainder, equal to two, corresponds to a consecutive negative sign and a positive sign: -, +;

- a remainder equal to three corresponds to two consecutive negative signs:

If the pattern of N characters corresponds to the value modulo 2 ^N of the sum of the coefficients in the sublist SE ₁ , then step C20 for entropy encoding the coefficients in the list Ε ₁ mentioned above is performed, except for the signs of the coefficient ε2 and the coefficient ε3, these signs being hidden in parity sums modulo 2 ^N coefficients.

If not, then step C10 is performed to modify at least one modified coefficient in sublist SE ₁ . Such modification is performed by the processing unit MTR_CO in FIG. 2 such that the modulus of the remainder 2 ^N of the sum of the modified coefficients in sublist SE ₁ reaches the value of each of the two characters to be hidden.

During the previously mentioned step C11, the processing module MTR_CO performs a corresponding modification of the list Ε ₁ . Thus, a modified list Em ₁ is obtained.

Then step C20 is performed to entropy-code the coefficients in the above list Em ₁ , except for the sign of the coefficient ε2 and the sign of the coefficient ε3, and these signs are hidden in the evenness of the modulo 2 ^N sum of the coefficients.

Detailed description of the decoding part

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

The decoding method according to the invention is in the form of an algorithm including steps D1-D12 shown in FIG. 3.

According to an embodiment of the invention, the decoding method according to the invention is embodied in a decoding apparatus or a DO decoder as shown in FIG. four.

During the preliminary step, which is not shown in FIG. 3, in the received data signal F, sections Β ₁ -B _Z are identified that were previously encoded by the CO encoder. In a preferred embodiment, said areas are blocks that are square in shape and all have the same size. Depending on the size of the image, which need not be an integer multiple of the size of the blocks, the last blocks on the left and the last blocks on the bottom need not be square. In an alternative embodiment, the blocks may be, for example, rectangular and/or may not be aligned with each other.

Each block or macroblock, furthermore, may itself be divided into sub-blocks, which may themselves be subdivided.

This identification is performed by the flow analysis software module EX_DO, as shown in FIG. four.

During step D1 shown in FIG. 3, the EX_DO module in FIG. 4 select as the current block B ₁ the first block B ₁ is to be decoded. Such a choice consists, for example, in placing the read pointer in signal F, at the beginning of the data elements of the first block B ₁ .

Then, each of the selected coding blocks is decoded.

In the example shown in FIG. 3, such decoding is applied sequentially to each of the coded blocks B ₁ to B _Z . The blocks are decoded, for example, in accordance with a raster scan, as is well known to a person skilled in the art.

The decoding according to the invention is implemented in the decoding software module MD_DO of the DO decoder as shown in FIG. four.

During step D2 shown in FIG. 3, the first entropy decoding of the first selected current block Β ₁ is performed. Such an operation is performed by the entropy decoding unit DE_DO shown in FIG. 4, for example, of the CABAS type. During this step, the DE_DO module performs entropy decoding of digital information items corresponding to the amplitude of each of the coded coefficients in list E ₁ or modified list Em ₁ . At this stage, only the signs of the coefficients in the list E ₁ or in the modified list Em ₁ are not decoded.

During step D3 shown in FIG. 3 define a set of characters that can be hidden during the previous entropy encoding step C20. Such step D3 is performed by the processing software module MTR_DO as shown in FIG. 4. Step D3 is similar to step C7 mentioned above for determining the number of characters to be hidden.

In a preferred embodiment, the number of hidden characters is one or zero. In addition, according to the preferred embodiment mentioned, the sign of the first non-zero coefficient is hidden. In the presented example, therefore, it represents the positive sign of the coefficient ε2=+9.

In an alternative embodiment, the number of hidden characters is zero, one, two, three, or more.

According to a preferred embodiment of step D3, during the first sub-step D31 shown in FIG. 3, a sublist containing the coefficients ε'1, ε'2, ..., ε'M, where Μ<L, which allow their modification when coding is determined from the mentioned list Ε ₁ or from the modified list Εm ₁ .

Such a determination is performed in the same way as in the above-mentioned encoding step C7.

Similar to the processing module MTR_CO mentioned above, the processing module MTR_DO is initially configured to modify:

- zero coefficient or coefficients located before the first non-zero coefficient,

- and for reasons of calculation complexity, a zero coefficient or coefficients located after the last non-zero coefficient.

In the example shown, at the end of sub-step D31, there is a sublist SEm ₁ such as SEm ₁ =(9, -6, 0, 0, 1, 0, -1, 2, 0, 0, 1). Therefore, eleven coefficients are obtained that could be modified.

During the next sub-step D32 shown in FIG. 3, the processing module MTR_DO proceeds to compare the number of coefficients that could be modified with a given threshold value TSIG. In the preferred embodiment, TSIG is 4.

If the number of coefficients that could be modified is less than the TSIG threshold, then during step D4 shown in FIG. 3 perform normal entropy decoding on all signs of the coefficients in list E ₁ . Such decoding is performed by the CABAC decoder, denoted by the reference value DE_DO in FIG. 4. To this end, the sign of each non-zero coefficient in the list E ₁ is subjected to entropy decoding.

If the number of coefficients that can be modified is greater than the threshold value TSIG, then during said step D4 a normal entropy decoding is performed on all signs of the coefficients in the list Εm ₁ except for the sign of the first non-zero coefficient ε2.

During step D5 shown in FIG. 3, the processing unit MTR_DO calculates a function value f that represents the coefficients in the sublist SEm to determine whether the calculated value is even or odd.

In the preferred embodiment, where only one character is hidden in signal F, the function f is the parity of the sum of the coefficients in the sublist SEm ₁ .

According to the convention used in the CO encoder, which is the same as in the DO decoder, an even value of the sum of the coefficients in the sublist SEm ₁ means that the sign of the first non-zero coefficient in the modified list Em ₁ is positive, while an odd value the value of the sum of coefficients in the sublist SEm ₁ means that the sign of the first non-zero coefficient in the modified list Em ₁ is negative.

In an exemplary embodiment in which SEm ₁ =(+9, -6, 0, 0, +1, 0, -1, +2, 0, 0, +1), the total sum of the coefficients is 6, and therefore, is even. Therefore, at the end of step D5, the processing unit MTR_DO deduces from it that the hidden sign of the first non-zero coefficient 2 is positive.

During step D6 shown in FIG. 3, and with all the reconstructed digital information elements during steps D2, D4 and D5, the quantized coefficients of block Bq ₁ are reconstructed in the given order. In the example shown, it is a reverse zigzag scan with respect to the zigzag scan performed during the encoding step C6 mentioned above. Such a step is performed by the read program module ML_DO as shown in FIG. 4. More specifically, the ML_DO module continues processing so as to include the coefficients of the list E ₁ (one-dimensional) in the block Bq ₁ (two-dimensional) using the reverse zigzag sweep order mentioned.

During step D7 shown in FIG. 3, the quantized residual block Bq ₁ is dequantized according to the conventional inverse quantization operation, which is the inverse quantization operation performed in the above encoding step C5, to obtain a decoded dequantized block BDq ₁ . Such a step is executed by the inverse quantization software module MDQ_DO as shown in FIG. four.

During step D8 shown in FIG. 3, an inverse transformation of the dequantized BDq ₁ block is performed, which is the inverse operation of the forward transformation performed in the encoding in step C4 mentioned above. Thus, a decoded residual block BDr ₁ is obtained. Such an operation is performed by the MTI_DO inverse transform software module as shown in FIG. four.

During step D9 shown in FIG. 3 perform predictive decoding of the current block B ₁ . Such predictive decoding is typically performed using known intra-frame and/or inter-frame prediction techniques, during which block B ₁ is predicted with respect to at least one previously decoded block. Such an operation is performed by the predictive decoding unit PRED_DO as shown in FIG. four.

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

During this step, decoding prediction is performed using the decoded syntax elements in the previous step, and in particular containing the type of prediction (interframe or intraframe), and, if necessary, the mode of intra frame prediction, the type of division of the block or macroblock, if the latter was divided , reference picture index, and motion vector, which are used in the inter prediction mode.

The predictive decoding step mentioned above generates a predictive block Bp ₁ .

During step D10 shown in FIG. 3, a decoded block BD ₁ is constructed by adding the decoded residual block BDr ₁ to the predicted block Bp ₁ . Such an operation is performed by the reconstruction software module MR_DO shown in FIG. four.

During step D11 shown in FIG. 3, the decoding module MD_DO checks whether the current decoded block is the last block identified in signal F.

If the current block is the last block in signal F, then during step D12 shown in FIG. 3, the decoding method ends.

If it is not, the next block B ₁ is selected as the block to be decoded, according to the order of the above raster scan, by repeating steps D1-D10, for 1 ≤ i ≤ Z.

Next, with reference to FIG. 3, another embodiment of the invention will be described.

This other embodiment differs from the previous one only in the number of latent coefficients, which is neither 0 nor N, where N is an integer such as N≥2.

To this end, the previously mentioned comparison sub-step D32 is replaced by the sub-step D32a represented by the dotted line in FIG. 3, during which the number of coefficients that could be modified is compared with several predetermined threshold values 0<TSIG_l<TSIG_2<TSIG_3, so that if the plurality of said coefficients is in the range from TSIG_N to TSIG_N+1, N characters can be hidden .

If the number of said coefficients is less than the first threshold value TSIG_1, then during the previously mentioned step D4, a normal entropy decoding of all signs of the coefficients in the list E ₁ is performed. To this end, the sign of each non-zero coefficient in the list E ₁ is subjected to entropy decoding.

If the number of said coefficients is between the threshold TSIG_N and TSIG_N+1, then during the previously mentioned step D4, a normal entropy decoding of all the signs of the coefficients in the list E ₁ is performed, except for the N corresponding signs of the first non-zero coefficients in the said modified list Em _1b the mentioned N signs hide.

In another embodiment, the processing module MTR_DO calculates, during step D5, the value of the function f, which is the remainder of module 2 ^N of the sum of the coefficients in the sublist SEm ₁ . In the proposed example, this is N=2.

The MTR_DO of the processing module therefore outputs from it the configuration of the two hidden characters that are assigned to each of the first two non-zero coefficients ε2 and ε3, respectively, in accordance with the convention used in encoding.

After the reconstruction of these two signs, perform steps D6 - D12, described above.

It goes without saying that the embodiments that have been described above have been presented solely as an indication and not limitation, and that a person skilled in the art can easily develop many modifications without thereby departing from the scope of the invention.

Thus, for example, in accordance with a simplified embodiment with respect to what is presented in Fig.1, the CO encoder could be configured to hide at least N' given characters, where N'≥1, instead of either zero , one or N specified characters. In this case, the comparison step C72 or C72a could be omitted. Accordingly, in accordance with a simplified embodiment of what is shown in FIG. 3, the DO decoder could be configured to reconstruct N' given characters instead of either zero, one, or N given characters. In this case, the comparison step D32 or D32a could be eliminated.

In addition, the determination criterion applied in the encoding step C72 or the decoding step D32 may be replaced by another type of criterion. To this end, instead of comparing the set of modified coefficients or the set of coefficients that have been modified with a threshold value, the MTR_CO or MTR_DO processing module could apply a selection criterion that is a function of the sum of the amplitudes of the coefficients that are or could be modified, respectively. , or the number of zeros present among the coefficients that are modified and that could be modified, respectively.

Claims (10)

1. A computer-implemented method, according to which a section (Bi) of an image divided into sections is decoded, comprising: - receiving a set of coefficients encoded by a CABAC (Context Adaptive Binary Arithmetic Coding) encoder; decoding (D2) said set of coefficients representing a chunk residual data block, the first non-zero coefficient not containing any sign value; - applying a function to a set of coefficients in order to generate sign data, wherein applying the function to a set of coefficients comprises: calculating the sum of non-zero coefficients from said set of coefficients; calculating the parity of said sum of coefficients; and moreover, this decoding method is characterized in that the sign is assigned to the first non-zero coefficient based on parity, and the even value of the mentioned sum of coefficients means that the sign of the first non-zero coefficient is positive, and the odd value of the said sum of coefficients means that the sign of the first non-zero coefficient is negative. 2. The method according to claim 1, characterized in that, in said decoding step (D2), first, entropy decoding of the first current block (Bi) that has been selected is performed, this being performed by an entropy decoding module (DE_DO) of the CABAC type. 3. The method according to claim 2, characterized in that the sign of each non-zero coefficient from the list (Ei) is decoded entropy. 4. Method according to claim 1, characterized in that said sections (Bi) are blocks which are square in shape and all have the same size.

Images