EP3183879A1

EP3183879A1 - Method for encoding a digital image, and associated decoding method, devices and computer programmes

Info

Publication number: EP3183879A1
Application number: EP15756178.8A
Authority: EP
Inventors: Félix Henry
Original assignee: Orange SA; B Com SAS
Current assignee: B Com SAS
Priority date: 2014-06-27
Filing date: 2015-06-16
Publication date: 2017-06-28
Also published as: JP2017521980A; US20170142444A1; WO2015197945A1; CN106797472A; KR20170026536A; FR3023112A1

Abstract

The invention concerns a method for encoding a digital image, said image being divided into a plurality of blocks of pixels processed in a defined order, said method comprising the following steps, implemented for a current block: -Predicting (E1) the values of the current block from at least one previously processed block, according to a prediction method chosen from a plurality of possible methods, -Calculating (E2) a residual block by subtracting the predicted values from the original values of the current block, -Obtaining (E3) a transformed residual block by applying a transform to the pixels of the residual block, said residual block comprising coefficients, -Selecting (E7) the signs of the coefficients to predict in the transformed residual block; -Predicting (E9) the selected signs in the current block from encoded neighbouring blocks that have been decoded; -Calculating (E10) a prediction indicator of the selected signs from the predictions of the selected signs and the original values of same, -Carrying out entropy encoding (E11) of the values of indicators obtained for the predicted signs. According to the invention, the method comprises a step of determining a context of a coefficient of the current residual block from a plurality of predefined contexts, and the sign of a coefficient of the transformed residual block is selected depending on a predefined score associated with the encoding context of said coefficient, said score being representative of a reliability level of the prediction of the sign.

Description

DIGITAL IMAGE ENCODING METHOD, DECODING METHOD, DEVICES, AND COMPUTER PROGRAMS

1. Field of the invention

The field of the invention is that of signal compression, in particular of a digital image or of a sequence of digital images, in which a prediction of a portion of the signal to be coded is made from a portion of the already coded signal.

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

images coming from the same camera and succeeding each other temporally (coding / decoding of 2D type),

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

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

- etc.

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

The invention may especially, but not exclusively, apply to video coding implemented in current AVC and HEVC video encoders and their extensions (MVC, 3D-AVC, MV-HEVC, 3D-HEVC, etc.), and to corresponding decoding.

The invention can also be applied to audio coding, for example implemented in current audio encoders (EVS, OPUS, MPEG-H, etc.) and their extensions and the corresponding decoding.

2. Presentation of the prior art

We consider a conventional compression scheme of a digital image, according to which the image is divided into blocks of pixels. A current block to be coded is predicted from a previously coded decoded block. A residual block is obtained by subtracting the original values from the predicted values. It is then transformed using a DCT type transformation (for "Discrete Cosine Transform", in English) or wavelets. The transformed coefficients are quantized and their amplitudes are subjected to entropic coding of the Huffmann or arithmetic type. Such coding obtains effective performances because, because of the transformation, the values of the amplitudes to be coded are for the most part zero. On the other hand, it does not apply to the values of the signs of the coefficients, whose + and - values are generally associated with equivalent appearance probabilities. Thus, the signs of the coefficients are coded by a bit 0 or 1.

The article by Koyama, J. et al, titled "Coefficient signal bit compression in video coding", and published in the proceedings of the "Picture Coding Symposium (PCS)" conference in May 2012, is known from a selection method. of coefficient signs of a residual block to predict. The proposed selection is based on a predetermined number of coefficients as a function of their amplitude and the size of the block from which they come. The selected signs are predicted. The predictions obtained are compared with the original values of the signs to determine the value of a prediction indicator, also called the residue of a predicted sign. This indicator can take two values, which are a first value representative of a correct prediction and a second value representative of an incorrect prediction. The rest of the signs are coded explicitly, without prediction.

An advantage of such a selection is to predict the value of a sign with a correct prediction probability greater than 50%, thus to allow the application of entropic coding to the values of the prediction indicators. This entropic coding encodes the sign information with an average bit rate less than one bit per sign, and thus makes it possible to increase the compression ratio.

3. Disadvantages of prior art

Nevertheless, this technique has at least two major disadvantages:

- Some signs whose correct prediction probability is close to 50% enter the selection of coefficients to predict. If this has no impact on the compression performance (no gain), there is an unnecessary increase in the number of calculations to be performed; - Some coefficients whose probability of correct prediction is high (greater than 50%) are not retained in the selection of coefficients to be predicted. There is then a loss of compression efficiency, because these coefficients could be used to further reduce the size of the coded signal.

4. Objectives of the invention

The invention improves the situation.

The invention particularly aims to overcome these disadvantages of the prior art. More specifically, an object of the invention is to propose a solution that more effectively selects the signs to be predicted.

Another objective of the invention is to propose a solution that is more efficient in compression.

5. Presentation of the invention

These objectives, as well as others that will appear later, are achieved by a method of coding a digital image, said image being divided into a plurality of blocks of pixels processed in a defined order, said method comprising the following steps, implemented for a current block: prediction of the values of the current block from at least one previously processed block according to a prediction mode chosen from among a plurality of predetermined modes,

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

- Obtaining a residue block transformed by applying a transform to pixels of the residue block, said transformed residue block comprising coefficients, selecting the signs of the coefficients to be predicted in the transformed residue block; Prediction of selected signs; Calculating a prediction indicator of the selected signs from the predictions of the selected signs and their original values, the indicator being adapted to take a value in a group comprising:

a first value representative of a correct prediction;

a second value representative of an incorrect prediction;

Entropy coding of the indicator values obtained for the predicted signs.

According to the invention, the method comprises a step of determining a context of a coefficient of the current residual block among a plurality of predetermined contexts, as a function of at least one characteristic belonging to a group comprising at least the size of the block , the coefficient amplitude, the frequency of the coefficient and the prediction mode of the current block, and the sign of a coefficient of a current block to be predicted is selected according to a predetermined score associated with the coding context of the coefficient , said score being representative of a reliability level of the prediction of the sign.

Thus, the invention is based on an entirely new and inventive approach to image coding, which consists in predicting the value of the signs of the coefficients of a residual block, when their prediction is considered sufficiently reliable. To do this, a coefficient is associated with a coding context for which it has previously established score values representative of a reliability level.

Unlike the prior art, which selects a predetermined number of signs to be predicted as a function of the amplitude of their coefficient and the size of the block from which they come, the invention bases its selection on a predetermined reliability of the prediction of signs. in a particular coding context. A coding context of a coefficient can be defined by a set of coding characteristics of the coefficient and the block to which it belongs. It is understood that the reliability of the prediction of the sign varies according to such characteristics.

With the invention, a plurality of contexts is thus defined, associated with levels of reliability that are distinct from one another. The characteristics taken into account in defining the coding context of the coefficient correspond to those for which an impact on the reliability of the prediction result has been observed. For example, it was found that the sign predictions were more reliable for a large block (for example, 16x16 or 32x32 pixels) than for a small block (for example 4x4 or 8x8 pixels). Similarly, the prediction of a sign is more reliable for a low frequency coefficient than a high frequency one.

The invention thus makes it possible to solve the technical problem of the cost of coding the signs of the coefficients of a residue block in a coding scheme of a digital image. Indeed, with the invention, it is first of all ensured that the values of the prediction indicators of the predicted signs that are actually encoded in the bit stream will take the representative value of a correct prediction in the vast majority of cases, in order to provide a favorable context for entropy coding and thus guarantee increased compression performance.

According to an advantageous characteristic of the invention, the score is predetermined during a preliminary step of estimating a probability of correct prediction of the sign in the context of the coefficient.

Thus, the score corresponds to the exact value of probability of correct prediction, which ensures a maximum level of performance of the compression.

For example, these probabilities are constructed for a coding context of the predetermined coefficient, before encoding and decoding, either by statistical accumulation on a set of signals representative of the signals to be coded, or by mathematical calculation from hypotheses on the distribution of the signs of the coefficients.

According to another aspect of the invention, the sign of a quantized coefficient is selected when said score is greater than a predetermined threshold. The higher the score, the more the prediction of the sign of the coefficient can be considered reliable. The selection is made by comparing the score of the coefficient with a threshold and the sign of the coefficient is selected, when the score of the coefficient is greater than this threshold.

One advantage of such a solution is that it is simple and economical in computing resources. According to another aspect of the invention, the score may take binary values, a first value being representative of a sign to be predicted, a second value being representative of a sign not to be predicted.

In this embodiment, the score is binary. The signs selected are those associated with a representative score of a level of reliability deemed sufficient. This reduces the complexity of the method, because there is more comparison of the score with a pre-established threshold, insofar as the score itself is indicative of the selection or non-selection of a sign given.

According to yet another aspect of the invention, the value of the threshold is predetermined. The threshold value is fixed. It is known to the coder and the decoder. For example, it is empirically determined by statistical analysis of entropy encoding performance applied to the predicted signs over a representative set of samples.

An advantage of this solution is that it is simple and easy to implement.

According to yet another aspect of the invention, the value of the threshold is adapted during coding according to characteristics of the coding.

The value of the threshold may vary during coding depending on the characteristics of the signal or the unit that performs the encoding.

An advantage of this solution is that it makes it possible to optimize the performance of the entropy coding over time.

According to a first variant, the value of the threshold is calculated by the coder and transmitted to the decoder in the bit stream.

According to a second variant, the value of the threshold is similarly calculated by the coder and the decoder.

According to yet another aspect of the invention, the entropy coding step of the prediction indicator value of the sign of a coefficient takes into account the predetermined score associated with the coding context of the coefficient.

In this way, we make the most of a priori knowledge of the level of reliability predicting the signs of the current block and improving the compression performance.

The method that has just been described in its various embodiments is advantageously implemented by a device for decoding a digital image according to the invention. Such a device comprises the following units:

Prediction of the values of the current block from at least one block previously processed according to a prediction mode selected from a plurality of predetermined modes,

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

Obtaining a residue block transformed by applying a transform to pixels of the residue block, said transformed residue block comprising coefficients, selecting the signs of the coefficients to be predicted in the transformed residue block; Prediction of selected signs

Calculating a prediction indicator of the selected signs from the predictions of the selected signs and their original values, the indicator being adapted to take a value in a group comprising:

a first value representative of a correct prediction;

a second value representative of an incorrect prediction;

Entropy coding of the indicator values obtained for the predicted signs. Said device is particular in that it comprises a unit for determining a context of a coefficient of the current residual block among a plurality of predetermined contexts, as a function of at least one characteristic belonging to a group comprising at least the size of the block, the amplitude of the coefficient, the frequency of the coefficient and the prediction mode of the current block, and in that the sign of a coefficient of the current residual block is selected according to a predetermined score associated with a context of coding said coefficient, said score being representative of a reliability level of the prediction of the sign.

Correlatively, the invention also relates to a method of decoding an image digital. Such a method comprises the following steps:

Prediction of the current block from at least one previously processed block and information relating to a prediction mode of the current block;

Entropic decoding of the coded amplitudes of the coefficients of a transformed residue block, extracted from the bit stream;

Selection of the signs of the coefficients to be predicted in the transformed residue block; Prediction of the values of the selected signs;

Decoding prediction indicator values of the selected signs from encoded data extracted from the bitstream, the indicator being adapted to take a value in a group comprising:

a first value representative of a correct prediction; a second value representative of an incorrect prediction; Calculating the decoded values of the selected signs from the decoded prediction indicator values;

- Reconstruction of the residual block coefficients from the decoded amplitudes, and decoded signs.

According to the invention, said method is particular in that it comprises a step of determining a context of a coefficient of the current residual block among a plurality of predetermined contexts, as a function of at least one characteristic belonging to a group comprising at least the size of the block, the magnitude of the coefficient, the frequency of the coefficient and the prediction mode of the current block, and in that the sign of a coefficient of the current residue block is selected according to a predetermined score associated with a coding context of said coefficient, said score being representative of a reliability level of the prediction of the sign.

It will be noted that the step of selecting the signs to be predicted is implemented in a similar way in the coding method and in the decoding method. As a result, the various embodiments or embodiments of the aforementioned coding method can be added independently or in combination with each other, to the steps of the decoding method as defined above. In particular, according to one aspect of the invention, the score is predetermined during a prior step of estimating a probability of correct prediction of the sign in the context of the coefficient.

According to another aspect of the invention, the sign of a coefficient is selected when said score is greater than a predetermined threshold.

According to yet another aspect of the invention, the score may take binary values, a first value being representative of a correct prediction of the sign of the quantized coefficient, a second value being representative of an incorrect prediction of the sign of the quantized coefficient. In yet another aspect, the threshold value is predetermined.

According to yet another aspect of the invention, the value of the threshold is adapted during decoding according to characteristics of the coding.

In yet another aspect, the entropy decoding step of the prediction indicator value of the sign of a coefficient takes into account the predetermined score associated with the coding context of the coefficient.

The decoding method that has just been described is advantageously implemented by a device for decoding a digital image according to the invention. Such a device is particular in that it comprises the following units:

Entropic decoding of the coded amplitudes of the coefficients of a transformed residue block extracted from the bit stream;

- Selection of the signs of the coefficients to be predicted in the transformed residue block;

Prediction of the values of the selected signs; Decoding prediction indicator values of the selected signs from encoded data extracted from the bitstream, the indicator being adapted to take a value in a group comprising:

a first value representative of a correct prediction;

a second value representative of an incorrect prediction; and

Calculating the decoded values of the selected signs from the decoded prediction indicator values;

According to the invention, the decoding device is particular in that it comprises a unit for determining a context of a coefficient of the current residual block among a plurality of predetermined contexts, as a function of at least one characteristic belonging to a group comprising at least the size of the block, the amplitude of the coefficient, the frequency of the coefficient and the prediction mode of the current block, and in that the sign of a coefficient of the current residual block is selected according to a predetermined score associated with a coding context of said coefficient, said score being representative of a reliability level of the prediction of the sign.

Correlatively, the invention relates to a user terminal.

Such a terminal is particular in that it comprises a device for encoding a digital image and a device for decoding a digital image according to the invention. The invention also relates to a computer program comprising instructions for implementing the steps of a method of encoding a digital image as described above, when this program is executed by a processor.

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

These programs can use any programming language. They can be downloaded from a communication network and / or recorded on a computer-readable medium. The invention finally relates to recording media, readable by a processor, integrated or not integrated with the encoding device of a digital image and the decoding device of a digital image according to the invention, possibly removable, respectively memorizing a computer program implementing an encoding method and a computer program implementing a decoding method, as described above.

6. List of figures

Other advantages and characteristics of the invention will emerge more clearly on reading the following description of a particular embodiment of the invention, given as a simple illustrative and nonlimiting example, and the appended drawings, among which: FIG. 1 schematically shows the steps of the method of encoding a digital image according to an exemplary embodiment of the invention; Figure 2 schematically shows a decoded current block of a decoded digital image; FIG. 3 schematically shows the steps of the method of decoding a digital image according to an exemplary embodiment of the invention; FIG. 4 shows an example of a simplified structure of a device for coding a digital image and a device for decoding a digital image according to one embodiment of the invention.

7. Description of a particular embodiment of the invention

The general principle of the invention is based on the selection of the signs of coefficients to be predicted as a function of a predetermined score representative of a reliability level of the prediction of the sign for a coding context associated with the coefficient.

In relation to FIG. 1, an original video consisting of a sequence of M images II, 12,... IM, with M nonzero integer is considered. The images are encoded by an encoder, the encoded data is inserted into a bit stream TB transmitted to a decoder via a communication network, or a compressed file FC, intended to be stored on a hard disk for example. The decoder extracts the coded data, then received and decoded by a decoder in a predefined order known from the encoder and the decoder, for example in the temporal order II, then 12, and then IM, this order being able to differ according to the mode of the decoder. production. When encoding an image Im, with m integer between 1 and M, it is subdivided into blocks of maximum size which can in turn be subdivided into smaller blocks. Each block will undergo an encoding or decoding operation consisting of a sequence of operations, including non-exhaustively a prediction, a residue calculation, a transformation, a quantization and an entropy coding. This sequence of operations will be detailed later.

During a step E0, the first block to be processed is selected as the current block C. For example, this is the first block (in lexicographic order). This block has NxN pixels.

It is assumed that there are L cuts in possible blocks numbered from 1 to L, and that the cut used on the block C is the cut number I. For example, there may be 4 possible cuts, in blocks of size 4x4, 8x8 , 16x16, and 32x32.

Moreover, the current decoded picture ID will be noted. Note that, in a video encoder, the ID image is (re) constructed in the encoder so that it can be used to predict the other pixels of the video. During a step E1, a prediction P of the original block C is determined. It is a prediction block constructed by known means, typically by motion compensation (block derived from a previously decoded reference image ), or intra prediction (block constructed from the decoded pixels immediately adjacent to the current block in the ID image). The prediction information related to P is encoded in the bit stream TB or compressed file FC. It is assumed here that there are K possible prediction modes m1, m2, ..., mK, with K nonzero integer, and that the prediction mode chosen for block C is the mode mk.

During a step E2, an original residue R is formed, by subtraction R = CP of the prediction P of the current block C to the current block C. During a step E3, the residue R is transformed into a residue block transformed, called RT, by a DCT transform or wavelet transform, all two known to those skilled in the art and in particular implemented in the JPEG standards for the DCT and JPEG2000 for the wavelet transform.

In E4, the transformed residue RT is quantized by conventional quantization means, for example scalar or vector, into a quantized residue block RQ. This quantized block RQ contains NxN coefficients. As known in the state of the art, these coefficients are scanned in a predetermined order so as to constitute a one-dimensional vector RQ [i], where the index i varies from 0 to N ² -1. The index i is called the frequency of the coefficient RQ [i]. Classically, these coefficients are scanned in increasing frequency order, for example along a zigzag path, which is known from the JPEG fixed image coding standard.

During a step E5, the amplitude information of the coefficients of the residual block RQ is encoded by entropy coding, for example according to a Huffman coding or arithmetic coding technique. By amplitude is meant here the absolute value of the coefficient. Amplitude coding means are for example described in the HEVC standard and in the article by Sole et al, entitled "Transform Coefficient Coding in HEVC", published in the journal IEEE Transactions on Circuits and Systems for Video Technology, Volume 22, Issue: 12, pp. 1765 - 1777, December 2012. Classically, we can code for each coefficient representative information because the coefficient is non-zero. Then, for each nonzero coefficient, one or more information relating to the amplitude is encoded. CA amplitudes are obtained.

During a step E6, we associate to each coefficient of the block RQ a context Cxj among a plurality J of predetermined contexts, with J nonzero integer. Such a context is defined by at least one coding characteristic of the coefficient or block from which it is derived. Advantageously, the following characteristics are considered:

- the size of the RQ block,

the amplitude of the quantized coefficient RQ [i],

the frequency of the coefficient or index i in the block RQ,

the mode of prediction of the current block mk among the K possible modes.

Indeed, the prediction of the sign is all the more reliable as the amplitude is high.

Similarly, it was found that when the block is larger in size, the frequency of the coefficient lower, the prediction is more reliable. Finally, it was found that prediction is more reliable when the current block is associated with intra prediction of a certain type.

Alternatively, other contexts are conceivable. Thus, it is possible to take into account the type of image in which the current block is located, for example of the Intra or Inter type, known from the HEVC standard, as a function of the energy of the predictor P, or depending on the total number of non-zero coefficients in the current block.

During a step E7, one selects the signs of the coefficients of the block RQ to predict, according to a predetermined score Sj, with j integer between 1 and J, for the context Cxj associated with the coefficient RQ [i] considered . Such a score Sj is representative of a reliability level of the sign of the coefficient RQ [i].

According to a first embodiment of the invention, the score Sj takes values in a predetermined set, for example from 0 to 10.

According to a second embodiment, the score is a simple binary indication, one of which indicates that the sign will be predicted, and the other that the sign will not be predicted.

According to a third embodiment of the invention, the scores Sj correspond to probabilities known a priori, dependent on the context Cxj associated with the coefficient RQ [i]. We have, in the encoder, a set of probabilities of correct detection of the signs of the coefficients RQ. For example, this set of probabilities is stored in memory.

These probabilities were constructed before encoding and decoding, either by statistical accumulation on a set of signals representative of the signals to be coded, or by mathematical calculation from hypotheses on the distribution of the signs of the coefficients. For a coefficient RQ [i] associated with the context Cxj, we can thus obtain the score Sj [i] by calculating the probability p [l] [mk] [i] [| RQ [i] | ] correct prediction of the sign of the coefficient RQ [i].

Advantageously, the signs to be predicted are selected by thresholding the scores with which they are associated. Thus, for each coefficient RQ [i] which has a sign (that is to say, which is not zero) and which is associated with a context Cxj of score Sj, the sign is predicted if and only if Sj > T, where T is a predetermined threshold, for example equal to 0.7. For example, the threshold T is known from the encoder and the decoder. According to one variant, the threshold T may be chosen during encoding and written in the compressed file or in the bit stream comprising the coded data representative of the digital image Im. For example, if the unit that performs the encoding does not have enough computing resources at a given time, it is possible to increase this threshold T so as to predict fewer signs, and therefore implement fewer calculations.

It would also be possible to vary the threshold T according to the content of the images to be encoded: an image containing a lot of content, such as large variations in brightness or large movements would use a high threshold, and an image with little content such as that low luminosity variations or few movements, would use a lower threshold T, so as to smooth the complexity or the memory necessary for the coding of each image.

In the exemplary embodiment of the invention of FIG. 1, the steps E6 and E7 for determining the context of the coefficients and the selection of the signs to be predicted are based on the values of the quantized coefficients of the transformed residue block. Note that the invention is not limited to this particular case, these steps can also be implemented before the quantization of the residual block coefficients.

During a step E8, the set of RQ signs that are not predicted are coded in a conventional manner. It is known in particular from the H EVC standard, in particular from the article by Sole et al., Already mentioned, the fact of transmitting each sign in the form of a binary element 0 or 1, with a convention associating one with the sign more and the other sign less.

During a step E9, the selected signs are predicted as "to be predicted" in the RQ block. This is carried out by means known to those skilled in the art, for example according to the technique described in the article by Ponomarenko et al., Entitled "Prediction of DCT signals in block-based lossy image compression", published in the Proceedings of the SPIE 6497 Conference, Image Processing: Algorithms and Systems V, 64970L, February 2007. In one embodiment of the invention, as many decoded blocks as combinatorics of the signs to be predicted are constructed. Each decoded version uses a different combination of the signs to predict. For example, suppose the RQ block equals {+8, +7, 0, -6, -3,0,0,2, -1,0,0,0,0,0,0,0}. Suppose also that the signs to predict are those of ¹ and 4 ^th coefficients (amplitude 8 and 6 respectively). We already know the signs of the 2 ^nd , 5 ^th , 8 ^th and 9 ^th coefficients, which were not to be predicted. In our example, there are two signs to predict, which can take the values {+, +}, {+, -}, {-, +} and {-, -}. We will build four virtual blocks RQVs:

RQVO = {+8, +7, 0, +6, -3,0,0,2, -1,0,0,0,0,0,0,0}

RQV1 = {+8, +7, 0, -6, -3,0,0,2, -1,0,0,0,0,0,0,0}

RQV2 = {-8, +7, 0, +6, -3,0,0,2, -1,0,0,0,0,0,0,0}

RQV3 = {-8, +7, 0, -6, -3,0,0,2, -1,0,0,0,0,0,0,0}

Each RQVs block is then decoded with the conventional means of dequantization and inverse transform, adding the predicted block P to produce S virtual decoded blocks BDVs. The likelihood of each of these blocks is tested with a likelihood criterion. The combination of signs corresponding to the virtual decoded block that maximizes the likelihood criterion is retained.

Advantageously, the likelihood criterion used is the minimization of the quadratic error along the boundary between the virtual decoded block and the previously decoded pixels.

In relation to Figure 2, there is shown a decoded image ID and DVs virtual decoded block of size NxN pixels of the image, where DVs (n, m) is the value of DVs pixel of the block located on the n ^'th row and m ^th column of the block. The broken line F represents the boundary between the decoded virtual block and the rest of the image (previously decoded). ID (k, l) is the value of the ID pixel located on the k ^th line and the i ^th column of the image, and (lin, col) are the coordinates of the DVs block (coordinates of the pixel at the top left of DVS) in the image ID. We consider a "Side Matching" operator SM (3, B)), with 3 current image and B current block, defined as follows: iV-l iV-1

5M (3, B) = V (5 (0, a) - 3 (Zin - 1, col + af + Y (β (α, O) - 3 (Zin + a, col - 1)) ² In FIG. 2, this amounts to forming the sum (x1-yl) ² + (x2-y2) ² + (x3-y3) ² + (x4-y4) ² + (x5-y4) ² + (x6-y5) ) ² + (x7-y6) ² + (x8-y7) ² .

We will determine the optimal virtual decoded DVopt block that minimizes this measure:

DV _opt = argmin SM _DV (lD, DV _s ) where ID represents the reconstructed image after decoding.

Alternatively, the likelihood criterion used is the minimization of the error with the predictor P. This consists in selecting the virtual decoded block which minimizes the error with the predictor P.

The virtual residue associated with the optimal virtual decoded block is thus identified. Suppose for example that DV _opt = DV3, the associated virtual block is then RQV3. We then consider the signs assigned to the identified virtual residue, in our example {-, -}. These signs become predicting signs prediction sign associated with the coefficient is ¹ - and the prediction associated with the sign of the 4 ^th coefficient is also -.

During a step E10, for each sign to be predicted, information is calculated representative of the difference between the prediction of the sign and the real value of the sign, called the IP prediction indicator or the sign's residual.

Thus, in the previous example, the sign prediction is {-, -} while the true signs are {+, -}.

By convention, the IP prediction indicator is set to 1 when the prediction is correct and to 0 when the prediction is incorrect.

During a step El i, the values of the IP prediction indicator for each sign to be predicted are coded by a known entropy coding technique, such as for example a Huffman coding, arithmetic coding or CABAC coding such as 'used in the HEVC standard. A CIP value of the coded prediction indicator is obtained.

According to the invention, since only the signs that are associated with a representative score of a sufficient level of reliability are predicted, the prediction indicator more often takes the value 1 than the value 0. This is exploited by entropy coding to reduce the size of the compressed signal. Advantageously, the entropy coding takes into account the score Sj associated with the predicted sign for coding the IP indicator. For example, in the embodiment of the invention in which the score has a value between 0 (low reliability of the prediction) and 10 (high reliability of the prediction), the entropic coding of the indicators is parameterized taking into account the score, so as to exploit the more or less uniform distribution of the indicators. For example, CABAC-type entrapic coding, known from the HEVC standard, is used by initializing the probabilities used in CABAC based on the predetermined scores.

During a step E12, the decoded block D corresponding to the block RQ is constructed by applying to the quantized residue RQ the dequantization and reverse transform steps (known per se). A decoded residue block RD is obtained. The predictor block P is added to RD to obtain the decoded block D.

During this step we also add the decoded block D to the reconstructed image ID. This makes it possible to have in the encoder a decoded version of the current image. This decoded version is used in particular during the step of constructing a prediction of the signs selected to be predicted.

During a step E13 the coded data, that is to say the amplitudes of the coefficients CA, the unpredictable signs coded CS, the indicators of the predicted signs CIP are inserted in the bit stream TB or in the compressed file FC . During a step E14, it is tested whether the current block C is the last block to be processed by the coding unit, given the order of travel defined above. If yes, the coding unit has finished processing. If not, the next step is the step of selecting the next block E0. This block becomes the current block to be processed, and the next step is the prediction step E1. In an exemplary embodiment of the invention, the context Cxj depends on the size of blocks I (out of 4 possible sizes, as previously described), of the intra mk prediction mode among 35 possible prediction modes (as described in the standard HEVC mentioned above), the frequency i (among 16, 64, 256 or 1024 possible frequencies, depending on the size of the blocks), and the amplitude | RQ [i] | (which can take 256 possible values when encoded on 8 bits). In this example, the number J of contexts used is equal to 35x (16 + 64 + 256 + 1024) x256 = 12185600. A preliminary examination on typical video sequences makes it possible to calculate a probability of correct detection of the sign for each context Cxj. This probability is the score Sj associated with each context Cxj, which makes it possible to select the signs to be predicted from a threshold of 0.7, as previously described. Thus, a compression gain of 1 to 2% is observed compared to the state of the art.

In relation to FIG. 3, the steps of the method of decoding a coded digital image according to an exemplary embodiment of the invention are now presented.

A TB bitstream or an FC compressed file produced by the encoding method according to the invention which has just been described is considered. One or the other encodes a video composed of a sequence of M digital images Im, with M nonzero integer and m integer between 1 and M. An image Im is subdivided into blocks of size NxN, with N nonzero integer and for example equal to 4, 8, 16 or 32 pixels.

The decoding method according to the invention comprises a step D0 of selecting a first block to be decoded D ', which is identical to the step E0 of selecting a first block to be coded presented in relation with FIG.

During a step D1, a prediction P 'of the block to be decoded D' is determined. The prediction information related to P 'is read in the bit stream or the compressed file and is decoded. This prediction information comprises a prediction mode mk of the block to be decoded C current.

According to one variant, the prediction mode can instead be totally inferred.

During a step D2, the amplitude information of a quantized residual block RQ 'corresponding to the block to be decoded D' are read in the bit stream or the compressed file and then decoded. At the end of this step, we therefore know the magnitudes of the coefficients of the quantized residual block RQ '[i], with i integer between 1 and NxN, but not yet the signs.

During a step D3, the coding contexts Cxj 'of the coefficients of the quantized residual block RQ' are determined from among a plurality of predetermined contexts. This step is identical to that of the coding method. During a step D4, the signs of the coefficients RQ '[i] to be predicted are selected. This step is identical to that implemented at coding. It associates with each coefficient RQ '[i] the coding context Cxj' previously determined and is based on a predetermined score Sj 'representative of a reliability level of a prediction of the sign for the context Cxj' coefficient considered.

During a step D5, the unpredicted signs are decoded with means adapted to those used during the coding. Typically, the decoding implemented is binary, for example entropic or Huffman. The unpredicted NPS 'decoded signs are obtained.

During a step D6, the signs of the selected coefficients are predicted. This step is identical to that implemented at coding. A list of RQV virtual residue blocks is thus obtained, similar to that of the RQV residue blocks obtained at the coder. Each RQV block is then decoded with the conventional means of dequantization and inverse transform, adds the predicted block P 'to produce S virtual decoded blocks BDV's. The likelihood of each of these blocks is tested with a likelihood criterion. The one whose combination of signs maximizes this criterion is selected.

In the previous example, it was the combination {-, -}.

During a step D7, the bitstream is extracted and the DIP values' of an IP prediction indicator of the predicted signs are decoded. It is an information representative of a difference between the prediction of a sign and the real value of this sign, that is to say of a sign residue. It can take the following values: - a value representative of a correct prediction; a value representative of an incorrect prediction.

During a step D8, the values of this IP indicator in the current quantized residual block RQV are used to correct, if necessary, the predicted values of the selected signs.

Thus, in the preceding example, the values of the prediction indicator present in the bit stream are representative of a first false prediction and a second right prediction. This allows us to decode the real signs of the coefficients of the current residue block {+, -}, and to reconstruct the complete decoded residue RQ '= {+8, +7, 0, -6, -3,0,0,2 , -1,0,0,0,0,0,0,0}. During a step D9, the block RQ 'is dequantized to obtain a dequantized block RT'. This is done by means adapted to the quantification used during the coding (scalar dequantization, vector dequantization ...)

During a step D10, the dequantized residue RT 'is subjected to a transform that is the inverse of that used for coding. The decoded residue R 'is then obtained.

During a step DU, the decoded residue R 'is added to the prediction P', to reconstruct the decoded block D '. This block D 'is integrated in the image ID being decoded.

During a step D12, it is tested whether the current block is the last block to be processed by the decoding unit, given the order of travel defined above. If yes, the coding unit has finished processing. If not, the next step is the step of selecting the next block DO.

During a step D13, one comes to select the next block to be processed by the decoding unit, following the path defined above. This block becomes the current block to be decoded, and the next step is the prediction step D1.

It will be noted that the invention which has just been described can be implemented by means of software and / or hardware components. In this context, the terms "module" and "entity", used in this document, may correspond either to a software component, or to a hardware component, or to a set of hardware and / or software components, capable of implementing perform the function (s) described for the module or entity concerned.

With reference to FIG. 4, an example of a simplified structure of a device 100 for encoding a digital image according to the invention is now presented. The device 100 implements the coding method according to the invention which has just been described in relation with FIG.

For example, the device 100 comprises a processing unit 110, equipped with a μΐ processor, and driven by a computer program Pg l 120, stored in a memory 130 and implementing the coding method according to the invention.

At initialization, the code instructions of the computer program Pgi 120 are for example loaded into a RAM before being executed by the processor of the processing unit 110. The processor of the processing unit 110 implements the steps of the method described above, according to the instructions of the computer program 120.

In this exemplary embodiment of the invention, the device 100 comprises at least one prediction unit PRED of the values of the current block from at least one block previously processed according to a prediction mode chosen from among a plurality of predetermined modes, a CALC unit for calculating a residue block by subtracting the predicted values from the original values of the current block, a TRANS unit for obtaining a residue block transformed by application of a transform to the pixels of the residue block, said transformed residue block comprising coefficients, a DET unit for determining a context of a coefficient of the current residual block among a plurality of predetermined contexts, as a function of at least one characteristic belonging to a group comprising at least the size of the block, the amplitude of the coefficient, the frequency of the coefficient and the prediction mode of the current block, a unit SEL of selection of the signs of the coefficients to be predicted e in the current block, the sign of a coefficient of the transformed residue block being selected according to a predetermined score associated with the coding context of said coefficient, said score being representative of a reliability level of the prediction of the sign, a PRED SIGNS unit for prediction of the selected signs in the current block from coded neighboring blocks decoded according to an error minimization criterion, an IP CALC unit for calculating a prediction indicator of the selected signs from the predictions of the selected signs and their original values, the flag being intended to take a value in a group comprising a first value representative of a correct prediction and a second value representative of an incorrect prediction, an IP coding unit of entropic coding of the values of indicators obtained for the predicted signs and an NPS COD unit coding the signs not predicted.

The device 100 further comprises a unit BD1 for storing the coding contexts of the coefficients and predetermined scores associated with each of these contexts. These units are driven by the μΐ processor of the processing unit 110. Advantageously, such a device 100 can be integrated with a user terminal TU. The device 100 is then arranged to cooperate at least with the following module of the terminal TU: a data transmission / reception module E / R, through which the bit stream TB or the compressed file FC is transmitted in a network telecommunications network, for example a wired network or a wireless network.

Still in relation to FIG. 4, an example of a simplified structure of a device 200 for decoding a digital image according to the invention is now presented. The device 200 implements the decoding method according to the invention which has just been described in relation to FIG.

For example, the device 200 comprises a processing unit 210, equipped with a processor μ2, and driven by a computer program Pg2 220, stored in a memory 230 and implementing the decoding method according to the invention. At initialization, the code instructions of the computer program Pg2 220 are for example loaded into a RAM before being executed by the processor of the processing unit 210. The processor of the processing unit 210 sets implement the steps of the method described above, according to the instructions of the computer program 220. In this embodiment of the invention, the device 200 comprises at least one prediction unit PRED 'of the current block from at least a previously processed block and information relating to a prediction mode of the current block, a DEC RES unit for entropic decoding of the coded amplitudes of the coefficients of a residue block extracted from the bit stream, said residual block having been obtained by subtraction of the values predicted from said at least one previously processed block and information relating to a prediction mode mk of the current block, to the original values of the current block, a e DET 'unit for determining a context of a coefficient of the current residual block among a plurality of predetermined contexts, as a function of at least one characteristic belonging to a group comprising at least the size of the block, the amplitude of the coefficient , the frequency of the coefficient and the prediction mode of the block current, a unit SEL 'for selecting the signs of the coefficients to be predicted in the transformed residue block, the sign of a coefficient of the current residue block being selected according to a predetermined score associated with the coding context of said coefficient, said score being Representative of a reliability level of the prediction of the sign, a DEC NPS 'unit for decoding unpredicted signs from the coded data extracted from the bitstream, a PRED' SIGNS unit for predicting the values of the signs selected from the signs of pixels of at least one decoded neighbor block of the current block according to an error minimization criterion, a DEC decoding unit of prediction indicator value values of the selected signs from the coded data extracted from the bit stream, the indicator being intended to take a value in a group comprising a first value representative of a correct prediction and a second value representative of an incorrect prediction, a CALC unit of the decoded values of the signs predicted from the values of decoded prediction indicators of the selected signs, a TRANSF-1 unit of inverse transformation of the amplitudes of coefficients of the transformed residue block RQ '[i ], a RECONST unit for reconstructing the coefficients of the residual block from decoded amplitudes and decoded signs.

The device 200 further comprises a unit BD2 for storing the coding contexts of the coefficients and predetermined scores associated with each of these contexts.

These units are driven by the processor μ2 of the processing unit 210. Advantageously, such a device 200 can be integrated with a user terminal TU. The device 200 is then arranged to cooperate at least with the following module of the terminal TU:

A data transmission / reception module E / R, through which the bit stream TB or the compressed file FC is received from a telecommunications network;

An image rendering device DISP, for example a terminal screen, through which the decoded digital image or the sequence of decoded images is returned to a user.

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

Claims

A method of encoding a digital image, said image (Im) being divided into a plurality of pixel blocks (C) processed in a defined order, said method comprising the following steps implemented for a current block:

Prediction (El) of the values of the current block from at least one previously processed block according to a prediction mode selected from a plurality of predetermined modes,

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

Obtaining (E3) a transformed residue block (RT) by applying a transform to pixels of the residue block, said transformed residue block comprising coefficients,

Selection (E7) of the signs of the coefficients to be predicted in the transformed residue block;

Prediction (E9) of the selected signs;

Computing (E10) a prediction indicator (IP) of the selected signs from the predictions of the selected signs and their original values, the indicator being adapted to take a value in a group comprising:

a first value representative of a correct prediction; a second value representative of an incorrect prediction;

- Entropy coding (El i) of the indicator values obtained for the predicted signs; said method being characterized by comprising a step (E6) of determining a context (Cxj) of a coefficient of the current residual block from a plurality of predetermined contexts, according to at least one characteristic belonging to a group comprising at least:

The size of the block,

The amplitude of the coefficient,

The frequency of the coefficient, The prediction mode of the current block; and in that the sign of a coefficient of the transformed residue block is selected according to a predetermined score (Sj) associated with the coding context (Cxj) of said coefficient, said score being representative of a reliability level of the prediction of the sign.

A method of coding a digital image according to claim 1, characterized in that the score is predetermined during a prior step of estimating a probability of correct prediction of the sign in the context of the coefficient.

3. Method for coding a digital image according to one of claims 1 and 2, characterized in that the sign of a quantized coefficient is selected when said score is greater than a predetermined threshold (T).

Method for coding a digital image according to one of Claims 1 to 3, characterized in that the score can take binary values, a first value being representative of a sign to be predicted, a second value being representative of a sign not to predict.

A method of coding a digital image according to claim 3, characterized in that the value of the threshold (T) is predetermined.

A method of coding a digital image according to claim 3, characterized in that the value of the threshold (T) is adapted during coding according to characteristics of the coding.

7. A method of coding a digital image according to one of the preceding claims, characterized in that the step (E10) of entropy coding of the prediction indicator value of the sign of a coefficient takes into account the predetermined score associated with the coding context of the coefficient.

A device (100) for encoding a digital image, said image being divided into a plurality of blocks of pixels processed in a defined order, said device comprising the following units, which can be implemented for a current block:

Prediction (PRED) of the values of the current block from at least one previously processed block according to a prediction mode selected from a plurality of predetermined modes;

Calculation (CALC) of a residual block by subtraction of the predicted values from the original values of the current block;

Obtaining (TRANS) a transformed residue block by applying a transform to pixels of the residue block, the transformed residue block comprising coefficients;

Selection (SEL) of the signs of the coefficients to be predicted in the transformed residue block;

Prediction (PRED SIGNS) of the selected signs;

Calculation (IP CALC) of a prediction indicator of the selected signs from the predictions of the selected signs and their original values, the indicator being intended to take a value in a group comprising:

- Entropy coding (IP COD) of the indicator values obtained for the predicted signs; said device being characterized in that it comprises a context determining unit (Cxj) of a coefficient of the current residual block among a plurality of predetermined contexts, as a function of at least one characteristic belonging to a group comprising at least one less:

The size of the block,

The amplitude of the coefficient, The frequency of the coefficient,

The prediction mode of the current block; and

in that the sign of a coefficient of the transformed residue block is selected according to the predetermined score associated with a coding context of said coefficient, said score being representative of a reliability level of the prediction of the sign.

A method of decoding a digital image from a bit stream comprising encoded data representative of said image, said image being divided into a plurality of processed blocks in a defined order, said method comprising the following steps implemented for a block, called current block: prediction (D1) of the current block from at least one block previously processed and information relating to a prediction mode of the current block;

Entropic decoding (D2) of coded amplitudes of the coefficients (RQ '[i]) of a transformed residue block extracted from the bit stream (TB);

Selection (D4) of the signs of the coefficients to be predicted in the transformed residue block;

Prediction (D6) of the values of the selected signs;

Decoding (D7) of prediction indicator values of the selected signs from encoded data extracted from the bit stream, the flag being for taking a value in a group comprising:

a first value representative of a correct prediction;

a second value representative of an incorrect prediction;

Calculating (D8) decoded values of the selected signs from the decoded prediction indicator values;

Reconstruction (DU) of the residual block coefficients from decoded amplitudes and decoded signs; said method being characterized in that it comprises a step (D3) of determining a context of a coefficient of the current residual block among a plurality of predetermined contexts, as a function of at least one characteristic belonging to a group comprising at least one less:

The size of the block, The amplitude of the coefficient,

The frequency of the coefficient,

The prediction mode of the current block; and in that the sign of a coefficient of the transformed residue block is selected according to a predetermined score associated with the encoding context of said coefficient, said score being representative of a reliability level of the prediction of the sign.

A device (200) for decoding a digital image (ID) from a bit stream (TB) comprising encoded data representative of said image, said image being divided into a plurality of processed blocks in a defined order, said device comprising the following units, which can be implemented for a block, referred to as a current block:

Prediction (PRED ') of the current block from at least one block previously processed and information relating to a prediction mode of the current block;

Entropic decoding (DEC RES) of the coded amplitudes of the coefficients of a transformed residue block extracted from the bit stream;

Selection (SEL ') of the signs of the coefficients to be predicted in the transformed residue block;

Prediction (PRED 'SIGNS) of the values of the selected signs;

Decoding (DEC IP) of prediction indicator values of the selected signs from encoded data extracted from the bit stream, the indicator being adapted to take a value in a group comprising:

a first value representative of a correct prediction;

a second value representative of an incorrect prediction;

Calculation (CALC) of the decoded values of the selected signs from the predicted values and the decoded prediction indicator values; said device being characterized in that it comprises a unit (DET ') for determining a context of a coefficient of the current residual block among a plurality of predetermined contexts, as a function of at least one characteristic belonging to a group comprising at least : The size of the block,

The amplitude of the coefficient,

The frequency of the coefficient,

A signal carrying a bit stream (TB) comprising coded data representative of a digital image, said digital image being divided into blocks of processed pixels in a defined order, characterized in that said encoded data is obtained according to the coding method according to the one of claims 1 to 7.

User terminal (TU) characterized in that it comprises a coding device of a digital image according to claim 8 and a device for decoding a digital image according to claim 10.

Computer program comprising instructions for implementing the method of encoding a digital image according to one of claims 1 to 7, when executed by a processor.

A computer program comprising instructions for implementing the method of decoding a digital image according to claim 9, when executed by a processor.