WO2013144197A1

WO2013144197A1 - Method and device for predicting an image block and corresponding methods and devices for coding and decoding

Info

Publication number: WO2013144197A1
Application number: PCT/EP2013/056505
Authority: WO
Inventors: Philippe Bordes; Pierre Andrivon; Philippe Salmon
Original assignee: Thomson Licensing
Priority date: 2012-03-29
Filing date: 2013-03-27
Publication date: 2013-10-03

Abstract

The method for predicting a block of a current image from at least one current reference image I_k comprises steps for: - determining (20), for the block, at least one implicit weighted parameter, and - determining (22) a predictor for the block from the current reference image using the implicit weighted parameter. The at least one implicit weighted parameter is determined from at least one weighted parameter of at least one reference image.

Description

METHOD AND DEVICE FOR PREDICTING AN IMAGE BLOCK AND CORRESPONDING METHODS AND DEVICES FOR CODING AND DECODING 1 . Domain of the invention

The invention relates to the general domain of image coding.

More specifically the invention relates to a device and method for predicting an image block. The invention also relates to the methods and devices for coding and decoding implementing the prediction method.

2. Prior art

The majority of coding/decoding methods for image sequences use prediction between images (inter-image prediction) or prediction in the image (intra-image prediction). Such a prediction is used to improve the compression of the image sequence. It consists in generating a prediction image for a current image having to be coded and in coding the difference, called the residual image, between this current image and the prediction image. The more the prediction image is correlated with the current image, the lower is the number of bits required to code the current image and therefore the more effective is the compression. However, the prediction loses its efficiency when there is a variation in luminosity between the images of the sequence or inside an image. Such a luminosity variation is for example due to a modification of illumination, to fade effects, to flashes, etc.

Methods for coding/decoding image sequences are known that take into account an overall variation in luminosity. Hence, within the framework of the standard H.264 described in the document ISO/IEC 14496-10, it is known to use a weighted prediction method in order to improve the compression in the case of a variation in luminosity. The weighted prediction parameters are explicitly transmitted per image slice and this is the case for each reference image. For example, H.264 enables the transmission of multiplicative parameters and offsets. Hereafter in the document and in the interests of simplicity, weighted prediction parameters will be referred to as weighted parameters. Illumination correction using weighted parameters is applied in the same way for all of the blocks of the slice that use a given reference image. This method can be costly in terms of bitrate as the weighted parameters are transmitted in the stream. Such parameters are called explicit parameters.

It is also known in the art to determine such weighted parameters from information available both at the coder and at the decoder. This method is less costly in terms of bitrate as the weighted parameters are not coded explicitly in the stream. Such parameters are called implicit parameters. In the case of the H.264 standard, implicit weighted parameters are calculated at the coder and at the decoder for a B type slice of a current image, that is to say a slice comprising blocks predicted from two reference images, according to the temporal distance separating the current image from one of the reference images and according to the temporal distance separating the two reference images. The temporal distances are determined from POCs (Picture Order Count). The implicit weighted parameters as defined in the H.264 standard are only for type B slices. Specifically, these implicit weighted parameters can not be used with type P slices, that is to say a slice comprising blocks predicted from a single reference image. In addition, in H.264, the implicit weighted parameters are only multiplicative parameters. Notably, the method for calculating implicit weighted parameters in H.264 does not enable offsets to be calculated, these are thus always considered as being null.

3. Summary of the invention

The purpose of the invention is to overcome at least one of the disadvantages of the prior art.

For this purpose, the invention relates to a method for predicting a block of a current image from at least one current reference image l_k comprising steps for:

- determining, for the block, at least one implicit weighted parameter, and

- determining a predictor for the block from the current reference image using the implicit weighted parameter. Advantageously, the at least one implicit weighted parameter is determined from at least one weighted parameter of at least one reference image. The method for predicting according to the invention advantageously enables weighted parameters to be determined independently of their nature, that is to say multiplicative or offset parameters. In addition, the method according to the invention also enables implicit weighted parameters to be determined for P type blocks.

According to a particular characteristic of the invention, the at least one weighted parameter of at least one reference image is an implicit or explicit weighted parameter.

According to another aspect of the invention, the at least one implicit weighted parameter is determined from at least one first weighted parameter of a first reference image with respect to the current reference image and from at least a second weighted parameter of a second reference image with respect to the current reference image.

According to a particular embodiment, the at least one implicit weighted parameter is equal to Tnk+(Tmk-Tnk)*T1 T0, where Tnk is the at least one first weighted parameter and TMK is the at least one second weighted parameter and where T1 is the temporal distance between the current image and the first reference image and TO is the temporal distance between the first and second reference images.

According to a particular characteristic of the invention, each of the at least one implicit weighted parameter, at least one first weighted parameter and at least one second weighted parameter is a multiplicative or an offset weighted parameter.

According to another particular embodiment, the at least one implicit weighted parameter is determined from at least a first weighted parameter of the current image l_k with respect to a third reference image Is and a second weighted parameter of the current image or a current slice to which the current block belongs with respect to the third reference image Is, the third reference image Is preceding the current reference image l_k in the decoding order.

According to a particular characteristic of the invention, the at least one implicit weighted parameter comprises a multiplicative weighted parameter Wck and an offset Ock calculated as follows:

Wck=Wcs Wks and Ock=Ocs-Wck*Oks where Wcs and Ocs are weighted parameters of the current image or the current slice with respect to the third reference image Is and where Wks and Oks are weighted parameters of the current reference image l_k with respect to the third reference image Is. The invention also relates to a method for coding an image comprising the prediction of a block of the image according to the prediction method of the invention, to determine a residue from the block and the predictor and to code the residue.

According to a particular characteristic of the invention, the coding method comprises the coding of an item of information of slice level indicating if the weighting parameters are explicit or implicit.

The invention also relates to a method for decoding a block of an image comprising the prediction of the block according to the prediction method of the invention, to decode a residue and reconstruct the block from the residue and the predictor.

The invention relates to a device for predicting a block of a current image from at least one current reference image l_k comprising:

- means for determining, for the block, at least one implicit weighted parameter Wck, and

- means for determining a predictor for the block from the current reference image using the implicit weighted parameter.

According to the invention, the at least one implicit weighted parameter is determined from at least one weighted parameter of at least one reference image.

The invention also relates to a device for coding an image comprising a device for predicting a block according to the invention, means for determining a residue from the block and the predictor and means for coding the residue. The invention also relates to a device for decoding an image comprising a device for predicting a block according to the invention, means for decoding a residue and means for reconstructing the block from the residue and the predictor.

4. List of figures

The invention will be better understood and illustrated by means of embodiments and advantageous implementations, by no means limiting, with reference to the figures in the appendix, wherein:

- figure 1 shows a method for prediction of an image block according to the invention, - figures 2 and 3 represent a current image Ic and reference images,

- figure 4 shows a variant for prediction of an image block according to the invention,

- figures 5 and 6 represent tables in which are stored weighted parameters of reference images and the current image or slice,

- figure 7 shows a coding device according to the invention, and

- figure 8 shows a decoding device according to the invention.

5. Detailed description of the invention

The invention relates to a method for coding an image belonging to an image sequence in a stream. Likewise the invention relates to a method for decoding such an image. According to the invention, the methods for coding and decoding implement a same method for predicting blocks.

The method for predicting a block Be of a current image Ic from a current reference image Ik is described in reference to figure 1 . The block Be belongs to a current slice of the image Ic. During a step 20, at least one implicit weighted parameter Wck is determined for the block Be with respect to the image Ik. The weighted parameter Wck is applied on the block of the image Ik to predict the block Be. According to a variant, two weighted parameters are determined: a multiplicative parameter Wck and an offset parameter Ock. According to the invention, the implicit weighted parameter Wck is determined from at least one weighted parameter of a reference image that can be Ik or another reference image. A first embodiment is described in reference to figure 2. In this example, there are two reference images In and Im that are respectively associated with weighted parameters Wnk and Wmk and possibly with Onk and Omk which are offsets. The weighted parameters Wnk, Wmk, Onk and Omk are associated with the images n and m with respect to the image Ik and are either implicit or explicit. In the case where k=n then Wnk is considered as being equal to 1 and Onk equal to 0, likewise if k=m, Wmk is considered as being equal to 1 and Omk equal to 0. The weighted parameter Wck is thus determined from two parameters Wnk and Wmk also taking into account the temporal position relative to images Ic, In and Im, and more particularly of the temporal distance between these images. Wck is for example determined according to the following equations:

Wck=Wnk+(Wmk-Wn k)*(Tc-Tn )/(Tm -Tn )

Likewise, the offset Ock is determined according to the following formula: Ock=On k+(Om k-On k)*(Tc-Tn )/(Tm-Tn ). Tc indicates for example the time at which the current image is displayed, Tn indicates the time at which the image In is displayed and Tm indicates the time at which the image Im is displayed. According to an embodiment variant the display times Tc, Tn and Tm are replaced by the POCs. According to an embodiment variant, to determine the weighted parameters (Wck, Ock), the use is favoured of parameters Wnk and Wmk that are explicit or equal to (1 ,0) with respect to the use of implicit parameters.

According to a variant described in reference to figure 3, the weighted parameter(s) (Wck, Ock) are determined from a weighted parameter Wks associated with the current reference image Ik with respect to a reference image Is and another weighted parameter Wcs associated with the current image or with part of the current image to which the current block belongs with respect to the image Is. The image Is is a reference image preceding the current reference image Ik in the decoding order. The weighted parameter Wck is thus determined from two parameters Wks and Wcs also taking into account the temporal distance between the images lc, In and Im according to the following equations:

Wck=Wcs/Wks

Likewise, the offset Ock is determined according to the following formula: Ock=Ocs-WckOks.

During a step 22, a predictor Bp is determined for the current block Be from implicit weighted parameter(s) determined in step 20. Generally the predictor is determined from at least one motion vector MV associated with the current block Be. According to a first embodiment, the block Be is a P type block. It is predicted with respect to the image Ik according to the following formula: Wck*Bk(MV)+Ock where Bk(MV) is a block of the image Ik determined from the motion vector MV.

According to a variant embodiment, the block Be is a B type block. It is predicted with respect to two reference images Ik1 and Ik2 according to the following formula: (Wck1 ^*Bk1 (MV1 )+Wck2^*Bk2(MV2)+Ock1 +Ock2)/2 where Bk1 (MV1 ) is a block of the image Ik1 determined from the motion vector MV1 and where Bk2(MV2) is a block of the image Ik2 determined from the motion vector MV2. Wck1 is determined with respect to the current reference image Ik1 and Wck2 is determined with respect to the current reference image Ik2. At the coder the motion vectors are determined by motion estimation. At the decoder the motion vectors are decoded from the stream.

An embodiment variant of the method for prediction is described in reference to figure 4. According to a variant, a table is filled that contains the implicit and explicit weighted parameters of reference images stored in memory during the coding of the current slice or image. In this table the implicit or explicit nature of weighted parameters is also specified. Such a table is shown in figures 5 and 6.

During a step 40, the reference images that are no longer used as reference images are deleted from the table.

During a step 42, the weighted parameters of the current image or slice that are explicit are added to the table. It should be noted that the table is filled with a multiplicative parameter equal to 1 and an offset parameter equal to 0 each time that an image refers to itself.

During a step 44, the weighted parameters of the current image or current slice that are implicit are determined and added to the table. In the example of figure 5, N(n)>1 where N(n) is the number of parameters stored in the table corresponding to the reference image In. In this example, N(n)=2, the weighted parameters corresponding to the reference image In being: (1 , 0) and e(Wmn, Omn). The weighted parameters Wen and Ocn of the current image or slice with respect to the reference image In are thus determined according to the following equations:

Wcn=1 +(Wmn-1 )^*(Tc-Tn)/(Tm-Tn)

and

Ocn=Omn^*(Tc-Tn)/(Tm-Tn). Tc, Tn and Tm represent either the display times of images or the POCs of these same images.

In the specific case of figure 6, N(m)=1 and there is n<m with N(n)>=2, the weighted parameters of the current image or slice relative to the reference image Ik are thus determined according to the following equations: Wcm=Wcn/Wmn and Ocm=Ocn-WcmOmn.

According to a variant, the weighted parameters stored in the k^th column are reordered so that the implicit parameters are found in the first column, followed by the parameters (1 ,0) and the explicit parameters. The explicit parameters and the parameters (1 ,0) determined with respect to the reference image Ik are thus preferred over the implicit parameters to determine the value of parameters of the current image or slice.

According to the coding method, the block Be of the current image lc is predicted according to steps 20 and 22 of the prediction method. A residue is determined from the block Be and the predictor Bp. For example, the residue is determined by subtracting pixel by pixel from the block Be the predictor Bp. The residue thus determined is coded in the stream F. The coding of the residue generally comprises a transformation (for example by a DCT) of the residue, a quantization of the transformed residue and entropy coding (for example VLC). Such steps are well known to those skilled in the art of video coders and are not further described hereafter. According to a variant embodiment, the coding method comprises the coding of an item of information at image slice level (for example in a slice header) indicating if the weighted parameters used to code the blocks of the slice are implicit or explicit. This information is for example coded in the slice header. According to a particular characteristic of the invention, the coding method also comprises an item of information at image level indicating that the weighted parameters used to code the blocks of the image are implicit or explicit. This information at slice level and at image level can co-exist. Thus, the information at image level can indicate that the image uses implicit weighted parameters and the information at slice level allows inside the image the use of explicit parameters for some slices. In the specific case of the H.264 standard, the information at image level is for example coded in a PPS (Picture Parameter Set). In general several images of a same sequence refer to the same PPS. The invention at slice level thus enables some of these images to be coded using explicit weighted parameters while those referring to a PPS indicate a use of implicit weighted parameters. Another solution is to duplicate the PPS, one indicating use of explicit weighted parameters and the other indicating use of implicit weighted parameters. However such as solution is more costly in terms of bitrate.

According to the decoding method, the block Be of the current image lc to be decoded is predicted according to steps 20 and 22 of the prediction method. A residue is decoded from the stream F. The decoding of the residue generally comprises an entropy decoding, an inverse quantization and an inverse transform. Such steps are well known to those skilled in the art of video coders and are not further described hereafter. The block Be is then reconstructed from the residue and the predictor Bp. For example, the residue is determined by adding pixel by pixel the residue and the predictor Bp.

According to a variant embodiment, the decoding method comprises the decoding of an item of information at image slice level indicating if the weighted parameters used to decode the blocks of the slice are implicit or explicit. This information is for example decoded in the slice header.

According to a particular characteristic of the invention, the decoding method also comprises the decoding of an item of information at image level indicating that the weighted parameters used to decode the blocks of the image are implicit or explicit. This information at slice level and at image level can co- exist. Thus, the information at image level can indicate that the image uses implicit weighted parameters and the information at slice level enables the use inside the image of explicit parameters for some slices. In the specific case of the H.264 standard, the information at image level is for example coded in a PPS (Picture Parameter Set).

The methods described above can be incorporated into a support that can be read and run by a computer.

The invention also relates to a coding device ENC described in reference to figure 7 and a decoding device DECOD described in reference to figure 8. In figures the modules shown are functional units, that may correspond or not to physically distinguishable units. For example, these modules or some of them can be grouped together in a single component or circuit, or constitute functions of the same software. Conversely, some modules may be composed of separate physical entities. Coding and decoding devices compatible with the invention are implemented according to a purely hardware realisation, for example in the form of a dedicated component (for example in an ASIC (Application Specific Integrated Circuit) or FPGA (Field-Programmable Gate Array) or VLSI (Very Large Scale Integration) or of several electronic components integrated into a device or even in a form of a mix of hardware elements and software elements.

The coding device ENC receives at input images I belonging to a sequence of images. Each image is divided into blocks of pixels each of which is associated with at least one item of image data, e.g. luminance and/or chrominance data. The coding device ENC notably implements a coding with temporal prediction. Only the modules of the coding device ENC relating to the coding by temporal prediction or INTER coding are shown in figure 7. Other modules not shown and known to those skilled in the art of video coders implement INTRA coding with or without spatial prediction. The coding device ENC notably comprises a calculation module ADD1 capable of subtracting a prediction block Bp pixel by pixel from a current block Be to generate a residue or residual block noted res. It also comprises a module TQ able to transform then quantize the residual block res into quantized data. The transform T is for example a DCT. The coding device ENC further comprises an entropy coding module COD able to code the quantized data into a binary stream F. It also comprises a module ITQ implementing the inverse operation to the module TQ. The ITQ module carries out an inverse quantization followed by an inverse transform. The module ITQ is connected to a calculation module ADD2 able to add pixel by pixel the block of data from the module ITQ and the prediction block Bp to generate a block of reconstructed image data that is stored in a memory DPB.

The coding device ENC also comprises a motion estimation module ME able to estimate at least one motion vector MV between the block Be and a current reference image Ik stored in the memory DPB. According to a variant the motion estimation is carried out between the current block Be and the source image corresponding to Ik in which case the memory DPB is not connected to the motion estimation module ME. The motion vector(s) in the case of the temporal prediction mode or INTER mode are transmitted to a motion compensation module MC. The motion vector or motion vectors are also transmitted to the entropy coding module COD to be coded in the stream F. The motion compensation module MC then determines from the reference image Ik previously reconstructed and stored in the memory DPB, a reference block Bref from the motion vector MV determined by the motion estimation module ME. The coding device ENC also comprises a processing module WP able to implement step 20 of the prediction method. It also comprises a calculation module MULT able to determine the predictor Bp according to step 22 of the prediction method. The decoding device DECOD is described in reference to figure 8. The decoding device DECOD receives at input a binary stream F representative of an image sequence. The stream F is for example transmitted by a coding device ENC. The decoding device DECOD comprises an entropy decoding module DEC able to generate decoded data, for example residues relating to the content of images. The decoding device DECOD also comprises a motion data reconstruction module. According to a first embodiment, the motion data reconstruction module is the entropy decoding module DEC that decodes a part of the stream F representative of motion vectors.

According to a variant not shown in figure 8, the motion data reconstruction module is a motion estimation module. This solution for reconstructing motion data via the decoding device DECOD is known as "template matching".

The decoded data relating to the content of images is then transmitted to a module ITQ able to carry out an inverse quantization followed by an inverse transform. The ITQ module is identical to the ITQ module of the coding device ENC having generated the stream F. The ITQ module is connected to a calculation module ADD3 able to add pixel by pixel the block from the ITQ module and the prediction block Bp to generate a block of reconstructed image data that are stored in a memory DPB. The decoding device DECOD also comprises a motion compensation module MC identical to the motion compensation module MC of the coding device ENC. If an INTER coding mode is decoded, the motion compensation module MC determines from a reference image Ik previously reconstructed and stored in the memory DPB, a reference block Bref from the motion vector MV decoded for the current block Be by the entropy decoding module DEC. The decoding device DEC also comprises a processing module WP able to implement step 20 of the prediction method. It also comprises a calculation module MULT able to determine the predictor Bp according to step 22 of the prediction method. In the case of a scalable coder, the coding of images of an enhancement layer uses images from a lower layer (for example: the base layer) to construct a prediction signal. The prediction signal can be constructed from images of a lower layer by using the prediction method described in reference to figures 1 and 2.

Claims

Claims 1 . A method for predicting a block of a current image from at least one current reference image Ik comprising steps for:

- determining (20), for said block, at least one implicit weighted parameter, and

- determining (22) a predictor for said block from said current reference image using said implicit weighted parameter,

characterized in that said at least one implicit weighted parameter is determined from at least one weighted parameter of at least one reference image.

2. The method for predicting according to claim 1 , wherein said at least one weighted parameter of at least one reference image is an implicit or explicit weighted parameter.

3. The method for predicting according to claim 1 or 2, wherein said at least one implicit weighted parameter is determined from at least one first weighted parameter (Wnk, Onk) of a first reference image with respect to said current reference image and from at least a second weighted parameter (Wmk, Omk) of a second reference image with respect to said current reference image.

4. The method for predicting according to claim 3, wherein said at least one implicit weighted parameter is equal to Tnk+(Tmk-Tnk)*T1 /TO, where Tnk is said at least one first weighted parameter and Tmk is said at least one second weighted parameter and where T1 is the temporal distance between said current image and said first reference image and TO is the temporal distance between said first and second reference images.

5. The method for predicting according to claim 4, wherein each of the at least one implicit weighted parameter, at least one first weighted parameter and at least one second weighted parameter is a multiplicative or an offset weighted parameter.

6. The method for predicting according to claim 1 or 2, wherein said at least one implicit weighted parameter is determined from at least a first weighted parameter (Wks, Oks) of said current reference image l_k with respect to a third reference image Is and a second weighted parameter (Wcs, Ocs) of said current image or a current slice to which said current block belongs with respect to said third reference image Is, said third reference image Is preceding said current reference image l_k in the decoding order.

7. The method for predicting according to claim 6, wherein said at least one implicit weighted parameter comprises a multiplicative weighted parameter Wck and an offset weighted parameter Ock calculated as follows:

Wck=Wcs/Wks and Ock=Ocs-Wck^*Oks where Wcs and Ocs are weighted parameters of said current image or of he current slice with respect to said third reference image Is and where Wks and Oks are weighted parameters of said current reference image l_k with respect to said third reference image Is.

8. A method for coding an image comprising: predicting a block of said image according to the method for predicting of one of claims 1 to 7, determining a residue from said block and said predictor and coding said residue.

9. The method for coding according to claim 8, comprising the coding of an item of information at slice level indicating if the weighted parameters are explicit or implicit.

10. A method for decoding a block of an image, comprising: predicting said block according to the method for predicting of one of claims 1 to 7, decoding a residue and reconstructing said block from said residue and said predictor.

1 1 . The method for decoding according to claim 10, comprising the decoding of an item of information at slice level indicating if the weighted parameters are explicit or implicit.

12. A device for predicting a block of a current image from at least one current reference image l_k comprising:

- means for determining (WP), for said block, at least one implicit weighted parameter,

- means for determining (MULT) a predictor for said block from said current reference image using said implicit weighted parameter,

13. The device for predicting according to claim 12, said device being configured to execute the steps of the method for predicting according to one of claims 1 to 7.

14. A device for coding an image comprising a device for predicting a block according to one of claims 12 or 13, means for determining a residue from said block and said predictor and means for coding said residue.

15. A device for decoding an image comprising a device for predicting a block according to one of claims 12 or 13, means for decoding a residue and means for reconstructing said block from said residue and said predictor.