WO2015185742A1

WO2015185742A1 - System for video sequence coding using enlarged coding zones

Info

Publication number: WO2015185742A1
Application number: PCT/EP2015/062612
Authority: WO
Inventors: Xavier Ducloux; Frédéric LANDAIS
Original assignee: France Brevets SAS
Current assignee: France Brevets SAS
Priority date: 2014-06-05
Filing date: 2015-06-05
Publication date: 2015-12-10
Anticipated expiration: 2016-12-05
Also published as: FR3022063B1; FR3022063A1

Abstract

The subject of the invention is a system for coding at least one video sequence S_e composed of a series of images in such a way as to produce an output stream F_s. The system comprises a module (100) for dividing the images of the sequence into N so-called initial partitions; a module (101) for determining N coding zones, a coding zone comprising a main partition as well as at least one complementary partition adjacent to the said main partition; N coding modules (102-104) carrying out an independent coding of the N coding zones so as to generate for each of it a multi-partition coding stream; a module (108) for extracting a main coding stream for each multi- partition coding stream, a main coding stream corresponding to the main partition of the coding zone being associated therewith; a module (109) for generating an output stream F_s composed of the main coding streams.

Description

SYSTEM FOR VIDEO SEQUENCE CODING USING ENLARGED CODING

ZONES

The invention relates to a system for video sequence coding using coding zones which are enlarged before coding and applies to the field of video compression.

The HEVC (High Efficiency Video Coding) compression standard makes it possible to obtain an improvement of the video compression rate with respect to the AVC (Advanced Video Coding) current standard of possibly as much as fifty percent, at the cost of greater computation complexity. Indeed, to retain equivalent video quality, the volume of computation is quadrupled. It should also be noted that the display frequency has doubled to reach sixty images per second.

Moreover, the UHD (Ultra High Definition) display format comprising four times as many pixels as the HD (High Definition) format is supported by the HEVC standard.

The innovations afforded in the new-generation coding standards such as HEVC give rise to significant computation volumes that have to be supported by the coding systems.

Spatial division of the image can be implemented so as to take account of the fact that the computation resources of the coding system (processors, memories) are limited. This technique consists in dividing the images into several partitions so as to distribute the computations over several computation modules, a computation module being responsible for the processing of at least one partition.

By way of example, video compression according to the HEVC standard includes division of the images into several independent rectangular partitions referred to as "tiles". The AVC standard also offered a possibility of dividing the image into several independent rectangular partitions referred to as "slices", but dividing was curbed to a division into several horizontal slices. In the subsequent description, the set of parameters making it possible to define the division partitions of an image is referred to as the image division.

Independent compression of the partitions of a divided image causes breaks in coding and in quality at the boundaries of the partitions. Indeed, the images of a video sequence are predominantly coded in inter mode (temporal coding) by seeking for each block of the image that block which best corresponds to it in a previously coded image. When an object is in motion, the object may pass from one partition of the divided image to another partition, which means that the Inter prediction will not find any like block within the partition. The edge of the partitions is of course affected more by this phenomenon. In this case, the only possible coding is the Intra coding (spatial coding) where the prediction is constructed on the basis of the pixels of the current image and not of the previously coded images. This Intra coding of the boundaries of the partitions in the predominantly Inter coded Inter images may render these boundaries visible.

Furthermore, these defects are accentuated by the discontinuity of the predictions, the break in the motion vector field, and the quantization levels which may be different from one coded partition to the next. This coding difference is strongly perceived by the human eye when the boundaries are fixed over time, thus degrading the perception of the user. Once the boundary is detected by the observer, the latter will thereafter search for it unconsciously and will find it easily. This may be particularly damaging when viewing high-quality video in the UHD (Ultra-High Definition) format. In this case, the viewer expects impeccable quality and the presence of visible boundaries cannot be accepted.

Several techniques belonging to the prior art make it possible to improve the coding of a partitioned video.

A first existing technique implements an exchange of pixels between the codings of the partitions. In this case, each item of coding hardware in charge of a partition exchanges the pixels neighbouring the boundaries with adjacent partitions. This exchange enables the coding decisions at the boundaries to be optimized and made consistent. This technique has the major drawback of being expensive in memory bandwidth since it requires the exchange of a large volume of data between the coders. Indeed, the coders exchange the decoded pixels, that is to say the result of the internal decoding.

A second existing technique relies on an improvement in the quantization at the boundaries. In order to decrease the coding defects at the edges of the partitions, the bitrate or the quantity of information allocated to these edges is increased. This makes it possible to lower the quantization and to improve the coding. However, by considering a constant bitrate for the image, the increase in bitrate on the edges of the partitions involves a decline in bitrate and in quality over the remainder of the image. An aim of the invention is notably to alleviate the aforementioned drawbacks.

For this purpose the subject of the invention is a system for coding at least one video sequence S_e composed of a series of images in such a way as to produce an output stream F_s, the said system comprising:

- a module for dividing the images of the sequence into N so- called initial partitions;

- a module for determining N coding zones, a coding zone comprising a main partition as well as at least one complementary partition adjacent to the said main partition; - N coding modules carrying out an independent coding of the N coding zones so as to generate for each of it a multi-partition coding stream;

- a module for extracting a main coding stream for each multi- partition coding stream, a main coding stream corresponding to the main partition of the coding zone being associated therewith; - a module for generating an output stream F_s composed of the main coding streams.

According to one aspect of the invention, the module for determining the coding zones transmits signalization data to each coding module so as to provide notification of the dimensions of the main partition and complementary partitions of which their coding zone is composed.

A signalling stream is for example integrated into the output stream F_s, this signalling stream integrating information for describing the main partitions such as their number, their position and their dimension.

The module for determining the coding zones transmits for example the index number of the main partition to be extracted from each multi-partition coding stream to the extraction module.

In one embodiment, the signalling stream complies with the HEVC standardized syntax.

By way of example, each coding module is implemented by a dedicated processor.

In one embodiment, a set of at least two coding modules is implemented by a processor that is dedicated thereto.

The subject of the invention is also a method for coding at least one video sequence composed of a series of images in such a way as to produce an output stream and comprising the following steps:

- dividing the images of the sequence into N so-called initial partitions;

- determining N coding zones, a coding zone comprising an initial partition as well as at least one complementary partition adjacent to the said initial partition;

- independently coding the coding zones so as to generate for each of it a multi-partition coding stream;

- extracting for each multi-partition coding stream a main coding stream corresponding to the initial partition of the coding zone associated with the said multi-partition coding stream; - generating an output stream composed of the main coding streams.

Other characteristics and advantages of the invention will become apparent with the aid of the description which follows given by way of nonlimiting illustration, and with regard to the appended drawings among which:

Figure 1 gives an exemplary coding system according to the invention;

Figure 2 presents main partitions as well as coding zones for an image of a video stream to be coded;

Figure 3 gives an exemplary coding zone in which a main partition is associated with eight complementary partitions; - Figures 4 and 5 give two examples of generalized coding zone;

Figure 6 illustrates in a simplified manner the coding method according to the invention. Figure 1 gives an exemplary coding system according to the invention.

This coding system processes a video sequence S_e composed of a series of images to produce an output stream F_s.

Accordingly, the system comprises several modules. In this description, a module designates a functional block. Stated otherwise, a module corresponds to the set of hardware and/or software means allowing the implementation of one or more functions. In this embodiment, the coding system receives as input a video sequence S_e composed of a series of images.

A module 100 is charged with dividing the images of the sequence into N partitions called main partitions. The function of another module 101 is to determine N coding zones. In this description, a coding zone corresponds to a set composed of one of the main partitions of the image with which at least one complementary partition is associated. A complementary partition is a partition adjacent to the main partition.

The data of the images 1 10-1 12 associated with each coding zone are transmitted respectively to N coding modules 102-104. These N coding modules 102-104 carry out an independent coding of the N coding zones so as to generate for each of it a multi-partition coding stream 130-132.

Signalling data 120-122 can also be transmitted from the module 101 to the coding modules 102-104 so as to notify them of the dimensions of the main and complementary partitions of which the coding zones are composed.

In a preferred embodiment, each coding module is implemented by a processor. It is also possible to use a given processor to implement several coding modes.

One of the fundamental advantages of the invention stems from the fact that the coding modules are not restricted to encoding solely the partition of the image which is associated with them. Each coding module encodes the set of partitions of which a coding zone is composed. Thus, the pixels of the main partition can be predicted using the pixels of the complementary coding zone or zones.

A module 108 is thereafter charged with the extraction of a main coding stream for each multi-partition coding stream. A coding stream is termed main when it corresponds to the main partition of the coding zone which is associated therewith. Thus, N main coding streams 140-142 are extracted 105-107 from the N multi-partition streams 130-132.

A module 109 thereafter generates an output stream F_s composed of the main coding streams 140-142. By way of example, the output stream F_s is constructed simply by multiplexing the N main coding streams 140-142. Furthermore, a signalling stream 150 is integrated into the stream F_s to provide the information relating to the main partitions, for example the number of partitions on the vertical and the horizontal of the image as well as the width and height of the partitions. In a preferred embodiment, the signalling stream 150 complies with the standardized syntax of the HEVC standard.

Figure 2 presents the main partitions as well as coding zones for an image of a video stream to be coded.

In this example, the width 21 1 of the image is 3840 pixels and the height 210 of the image is 21 60 pixels, thus corresponding to a video in the Ultra High Definition 4K (UHD 4K) forma. The image is divided up into nine so-called main partitions. Six main partitions 200-202, 206-208 have a dimension of 760 pixels by 1280 pixels and three main partitions 203-205 have a dimension of 640 pixels by 1280 pixels.

Two coding zones 220, 221 are also represented. According to the invention, there are as many coding zones as main partitions. Thus, in this example, the system and/or the method according to the invention are implemented using nine coding zones, even if here only two of them 220, 221 are represented.

It is apparent that the coding zone 221 is of greater dimension than the main partition 204 associated therewith. More precisely, in this example, the boundaries of the coding zone 221 correspond to the extended boundaries of the initial partition 222, 223 of 128 pixels. This corresponds respectively to extensions of twenty percent and forty percent of the horizontal and vertical dimensions of the initial partition 204. The greater these extensions, the better the tracking of the motion of objects from one main partition to the next. This consequently implies that the boundaries between partitions will be less visible for an observer of the video stream after decoding. Figure 3 gives an exemplary coding zone in which a main partition is associated with eight complementary partitions. These nine complementary partitions 301 -308 are adjacent to the main partition 300. Within the framework of this invention, the word adjacent signifies that the complementary partitions are in the immediate neighbourhood of the main partition. The main partition as well as the complementary partitions are encoded by a coding module generating a multi-partition stream comprising in fact nine coding streams. The extraction module 108 extracts the coding stream corresponding to the main partition. Advantageously, the inter coding of the coding zone by a coding module of the system according to the invention allows the main coding stream to take account of the zones of the image adjoining the main partition in a previous image (inter coding) of the video stream to be encoded. Thus, the boundaries between the main partitions of the images of the decoded video stream become difficult for observers to locate.

The person skilled in the art will deduce without effort that other configurations may be envisaged by varying the number, the position and the shape of the complementary zones. Figures 4 and 5 give two examples of generalized coding zone. It should be noted that these two examples are illustrative and are not supported by the AVC or HEVC compression standards. In particular, the HEVC standard implies that a coding zone (partitioned or not) must be rectangular and that the coding zone is divided up into a number of partitions which is equal to the product of the number of columns times the number of rows.

Figure 6 illustrates in a simplified manner the coding method according to the invention. The coding relates to at least one video sequence S_e composed of a series of images and the application of the method makes it possible to produce an output stream F_s.

The method comprises a step 600 of dividing the images of the sequence into N so-called initial partitions. The method also comprises a step 601 of determining N coding zones, a coding zone comprising an initial partition as well as at least one complementary partition adjacent to the said initial partition.

The method also comprises a step 602 of independent coding for each coding zone so as to generate a multi-partition coding stream for each of it.

The method also comprises 603 a step of extracting for each multi- partition coding stream a main coding stream corresponding to the initial partition of the coding zone associated with the said multi-partition coding stream.

Finally, the method comprises 604 a step of generating an output stream composed of the main coding streams.

Claims

System for coding at least one video sequence S_e composed of a series of images in such a way as to produce an output stream F_s, the said system comprising:

- a module (100) for dividing the images of the video sequence into N so-called initial partitions;

- a module (101 ) for determining N coding zones, a coding zone comprising a main partition as well as at least one complementary partition adjacent to the said main partition ;

- N coding modules (1 02-104) carrying out an independent coding of the N coding zones so as to generate for each of them a multi-partition coding stream ;

- a module (1 08) for extracting a main coding stream for each multi-partition coding stream, a main coding stream corresponding to the main partition of the coding zone being associated therewith ;

- a module (109) for generating an output stream F_s composed of the main coding streams.

The coding system according to Claim 1 , in which the module (101 ) for determining the coding zones transmits signalling data (120-122) to each coding module (102-104) so as to provide notification of the dimensions of the main partition and complementary partitions of which their coding zone is composed.

The coding system according to one of the preceding claims, in which a signalling stream (150) is integrated into the output stream F_s, this signalling stream integrating information for describing the main partitions such as their number, their position and their dimension. The coding system according to Claim 3, in which the module (101 ) for determining the coding zones transmits the index number of the main partition to be extracted from each multi-partition coding stream to the extraction module (108).

The coding system according to Claim 3, in which the signalling stream 150 complies with the HEVC standardized syntax.

The coding system according to one of the preceding claims, in which each coding module is implemented by a dedicated processor.

The coding system according to one of Claims 1 to 5, in which a set of at least two coding modules is implemented by a processor that is dedicated thereto.

Method for coding at least one video sequence composed of a series of images in such a way as to produce an output stream and comprising the following steps:

- dividing (600) the images of the video sequence into N so- called initial partitions;

- determining (601 ) N coding zones, a coding zone comprising an initial partition as well as at least one complementary partition adjacent to the said initial partition;

- independently coding (602) the coding zones so as to generate for each of them a multi-partition coding stream;

- extracting (603) for each multi-partition coding stream a main coding stream corresponding to the initial partition of the coding zone associated with the said multi-partition coding stream;

- generating (604) an output stream composed of the main coding streams.