WO2014205717A1

WO2014205717A1 - Video decoding method and decoder

Info

Publication number: WO2014205717A1
Application number: PCT/CN2013/078124
Authority: WO
Inventors: 董胜富; 王振宇; 王荣刚; 高文; 王文敏; 李英
Original assignee: 北京大学深圳研究生院
Priority date: 2013-06-27
Filing date: 2013-06-27
Publication date: 2014-12-31
Also published as: CN105264887A

Abstract

Disclosed are a video decoding method and a decoder. An entropy decoding module performs entropy decoding on a code stream corresponding to an input image frame, and at the same time, the entropy decoding module distributes row data obtained through entropy decoding to an idle rebuilding module one by one in a predetermined order according to idle state information about a rebuilding module; and the rebuilding module rebuilds rows in the image frame according to the received row data. In this way, when the entropy decoding module performs entropy decoding on a row in an image frame, at the same time, the rebuilding module can rebuild a previous row, thereby improving the video decoding efficiency and improving the video parallel decoding performance; and the rebuilding of rows in the image frame also can be performed in parallel among the rebuilding modules, thereby further improving the video decoding efficiency and further improving the video parallel decoding performance.

Description

Video decoding method and decoder

Technical field

The present application relates to the field of data processing, and in particular, to a video decoding method and a decoder.

Background technique

With the popularity of video services, video of various resolutions has been used more and more, especially video with high-definition resolution and ultra-high-definition resolution has gradually entered the network video service, correspondingly, high-definition resolution and ultra-high definition resolution. Rate video real-time decoding technology has become a key technology to improve the quality of video services.

At present, the H.264 or AVS standard has become the mainstream video compression coding standard currently applied on the market. Compared with previous standards such as MPEG2, the H.264 or AVS standard has greatly improved the complexity of codec operation, and the real-time decoding of HD resolution and ultra-high definition resolution has become a popular HD resolution and ultra-high definition resolution video service. The bottleneck.

At present, mainstream PC configurations have been processed using dual-core, quad-core or even eight-core CPUs, and mobile terminals are increasingly using multi-core ARM chips for processing. Although hardware decoding chips suitable for different video coding standards are integrated in PCs and mobile terminals, video codec technology is constantly updated and developed due to the diversification of coding formats, and the support of unified hardware decoding chips for video is limited. Sex, and it is difficult to adapt to the new video codec standards, making it difficult to quickly apply and popularize new video codec standards. Software-based video decoders have become the focus of current video decoder optimization. Software-based video decoders have extensively utilized CPU parallel computing capabilities for parallel video decoding. Typical parallel video decoding includes GOP level parallelism, frame level parallelism, and stripe parallelism. Since GOP-level parallel and frame-level parallel decoding have high delay characteristics, stripe parallel decoding is widely used in video decoding.

The existing stripe parallel decoding mainly performs serial entropy decoding and reconstruction on all macroblocks in the stripe, that is, only the entropy decoding and reconstruction of the previous macroblock in the stripe can complete the next macroblock. Entropy decoding and reconstruction. The serial execution of such strips still has a problem of high delay and low processing efficiency, which reduces video decoding performance.

Summary of the invention

The present application provides a video decoding method and decoder to improve video parallel decoding performance.

According to the first aspect of the application, the application provides a video decoding method, including:

The entropy decoding module performs entropy decoding on the code stream corresponding to the input image frame, and the entropy decoding module distributes the entropy decoded row data to the idle reconstruction module one by one according to the idle state information of the reconstruction module;

The reconstruction module reconstructs the row in the image frame according to the received row data, and sends a row reconstruction end message to the entropy decoding module when the row reconstruction is completed;

The entropy decoding module detects whether the reconstruction module completes reconstruction of the last row in the image frame according to the row reconstruction end message, and if so, entropy decodes the code stream corresponding to the next input image frame.

According to a second aspect of the present application, the application provides a video decoder, including an entropy decoding module and a reconstruction module, where:

An entropy decoding module, configured to perform entropy decoding on a code stream corresponding to the input image frame, and, according to the idle state information of the reconstruction module, distribute the entropy decoded row data to the idle reconstruction module one by one in a predetermined order;

a reconstruction module, configured to reconstruct a row in the image frame according to the received row data, and send a row reconstruction end message to the entropy decoding module to trigger the entropy decoding module to detect whether the last line of the image frame is completed when the row reconstruction is completed. Reconstruction and corresponding entropy decoding of the code stream corresponding to the next input image frame.

The beneficial effects of the application are:

By providing a video decoding method and a decoder, the entropy decoding module performs entropy decoding on the code stream corresponding to the input image frame, and at the same time, the entropy decoding module decodes the entropy decoded line data in a predetermined order according to the idle state information of the reconstruction module. The reconstruction module is distributed to the idle reconstruction module one by one; the reconstruction module reconstructs the rows in the image frame according to the received row data, and after completing the reconstruction of the row, the entropy decoding module determines whether the reconstruction of the last row of the image frame is completed, and if so, the next The code stream corresponding to an input image frame is entropy decoded. In this way, the entropy decoding module performs the entropy decoding of one row in the image frame, and the reconstruction module can reconstruct the previous row, which improves the video decoding efficiency and improves the video parallel decoding performance; the reconstruction module can also execute the image frame in parallel. The reconstruction further improves the video decoding efficiency and further improves the video parallel decoding performance.

DRAWINGS

1 is a flowchart of a video decoding method according to Embodiment 1 of the present application;

2 is a structural diagram of a video decoder according to Embodiment 2 of the present application;

3 is a flowchart of processing of an entropy decoding module and a reconstruction module in Embodiment 1 of the present application;

4 is a specific flowchart of step 102 in the first embodiment of the present application;

FIG. 5 is a schematic diagram showing the relationship between a current block and other blocks in an image frame according to Embodiment 1 of the present application;

FIG. 6 is a structural diagram of a reconstruction module 202 according to Embodiment 1 of the present application;

FIG. 7 is a specific flowchart of step 404 in the second embodiment of the present application;

FIG. 8 is a structural diagram of a reconstruction module 202 in Embodiment 2 of the present application;

FIG. 9 is a specific flowchart added in Embodiment 3 of the present application;

FIG. 10 is a structural diagram of an entropy decoding module 201 in Embodiment 3 of the present application.

detailed description

The present application will be further described in detail below with reference to the accompanying drawings.

Embodiment 1:

Referring to FIG. 1, the video decoding method of this embodiment is mainly based on a video decoder as shown in FIG. 2. In the video decoder, the shared memory stores data required for video decoding, and the data mainly includes a code stream corresponding to the image frame to be decoded, and various status information, such as idle state information and reconstruction state information described below. At the same time, the control elements in the video decoder provide the corresponding threads, and each thread can access the data stored in the shared memory. The entropy encoding module 201 and the reconstruction module 202 in the video decoder are each driven by independent threads. Based on this, the above video decoding method mainly includes the following processes:

Step 101: The entropy decoding module performs entropy decoding on the code stream corresponding to the input image frame. At the same time, the entropy decoding module distributes the entropy decoded row data to the idle reconstruction module one by one according to the idle state information of the reconstruction module. Specifically, the idle state information is used to indicate whether the re-establishment module is in an active state or an idle state, and may include an identifier of the re-establishment module and a symbol indicating whether it is idle, etc., and the entropy coding module may find an idle re-establishment module according to the idle state information. . The predetermined order is generally a code stream input sequence, and the previous line data of the code stream input sequence is preferentially sent to the reconstruction module for processing. Each reconstruction module processes only one row of data. As shown in FIG. 3, the entropy decoding module obtains the row data of the first row after completing the entropy decoding of the first row, and then the entropy decoding module sends the row data of the first row to the idle first reconstruction module. The first reconstruction module performs reconstruction of the first row according to the row data of the first row. At the same time that the first reconstruction module performs reconstruction, the entropy decoding module can perform entropy decoding on the second row to obtain row data of the second row. Then, the entropy decoding module sends the row data of the second row to the idle second reconstruction module, and the second reconstruction module performs the reconstruction of the second row according to the row data of the second row, and so on.

Step 102: The reconstruction module reconstructs the row in the image frame according to the received row data, and sends a row reconstruction end message to the entropy decoding module when the row reconstruction is completed. Specifically, step 102 may include a specific process as shown in FIG. 4:

In step 401, the reconstruction module extracts block data from the row data. Specifically, the block data is sequentially combined to form row data, and has a block identifier. Based on the block identification, the reconstruction module can identify the block data and proceed with the processing of the subsequent step 402.

Step 402: The reconstruction module detects whether the reconstruction of the block has a reconstruction condition. If yes, step 403 is performed. Otherwise, it waits for a certain time, and after waiting, proceeds to step 402. Specifically, the reconstruction condition may be an intra prediction condition or an inter prediction condition. The intra prediction condition refers to: entropy decoding and reconstruction have been completed in the left block, the upper left block, the upper block, and the upper right block of the current block in the image frame to which the current block belongs. The inter-frame prediction condition refers to: on the one hand, in the image frame to which the current block belongs, the left block, the upper left block, the upper block, and the upper right block of the current block have completed entropy decoding and reconstruction, and on the other hand, the reference to which the current block motion depends In the image frame, all blocks in the motion search range have been entropy decoded and reconstructed. The reconstruction module can access the shared memory, and the reconstruction state information of the block stored in the shared memory described below is used as the detection basis.

In step 403, the reconstruction module reconstructs the blocks in the row according to the block data. Specifically, the reconstruction module may sequentially perform intra prediction or inter prediction, inverse scanning of transform coefficients, inverse quantization, inverse transform, prediction block and residual block superposition according to block data, thereby obtaining a reconstructed image of the block.

Step 404: After the reconstruction of the current block is completed, the reconstruction module updates the reconstruction state information of the current block as a basis for detecting whether the reconstruction of the unprocessed block has a reconstruction condition. Referring to FIG. 5, when performing intra prediction on the current block marked with *, the left block, the upper left block, the upper block, and the upper right block that need to satisfy the # mark have completed entropy decoding and reconstruction, so that only the left block is updated. The reconstruction state information of the upper left block, the upper block, and the upper right block, the reconstruction module can detect whether the current block marked with * has a reconstruction condition.

Step 103: The entropy decoding module detects whether the reconstruction module completes reconstruction of the last row in the image frame according to the row reconstruction end message, and if yes, proceeds to step 101 to entropy decode the code stream corresponding to the next input image frame.

Accordingly, in the video decoder of this embodiment as shown in FIG. 2:

Entropy decoding module 201, for entropy decoding the code stream corresponding to the input image frame, and according to the idle state information of the reconstruction module 201, the entropy decoded row data is distributed to the idle reconstruction module 201 one by one in a predetermined order;

The reconstruction module 202 is configured to reconstruct a row in the image frame according to the received row data, and send a row reconstruction end message to the entropy decoding module 201 to trigger the entropy decoding module 201 to detect whether the image is completed when the row reconstruction is completed. The reconstruction of the last line of the frame corresponds to entropy decoding of the code stream corresponding to the next input image frame. Specifically, the reconstruction module 202 can include the structure as shown in FIG. 6:

a block reconstruction module 601, configured to extract block data from the row data, and detect whether the reconstruction of the block has a reconstruction condition, and if yes, reconstruct the block in the row according to the block data;

The progress update module 602 is configured to update the reconstruction state information of the current block after the block reconstruction module 601 completes the reconstruction of the current block, as the block reconstruction module 601 detects whether the reconstruction of the unprocessed block has a basis for the reconstruction condition.

By implementing the embodiment, the entropy decoding module performs entropy decoding on the code stream corresponding to the input image frame, and at the same time, the entropy decoding module distributes the entropy decoded row data to the idle one by one according to the idle state information of the reconstruction module. Reconstruction module; the reconstruction module reconstructs the rows in the image frame according to the received row data, and after completing the reconstruction of the row, the entropy decoding module determines whether the reconstruction of the last row of the image frame is completed, and if so, the image frame of the next input The corresponding code stream is entropy decoded. In this way, the entropy decoding module performs the entropy decoding of one row in the image frame, and the reconstruction module can reconstruct the previous row, which improves the video decoding efficiency and improves the video parallel decoding performance; the reconstruction module can also execute the image frame in parallel. The reconstruction further improves the video decoding efficiency and further improves the video parallel decoding performance.

Embodiment 2:

This embodiment adds the following content on the basis of the first embodiment:

In the video decoding method, after step 404, the process shown in FIG. 7 is also performed:

Step 701: The reconstruction module determines whether all the blocks in the row of the block are completely reconstructed. If yes, step 702 is performed; otherwise, step 401 is performed. Specifically, the reconstruction module detects that the input row data is processed, indicating that the entire row has been reconstructed.

In step 702, the reconstruction module performs filtering and interpolation processing on the row of the block.

Correspondingly, the reconstruction module 202 in the video decoder of this embodiment further includes the structure shown in FIG. 8:

The determining module 801 is configured to determine whether all blocks in the row of the block are completely reconstructed;

The filtering and interpolation module 802 is configured to filter and interpolate the row of the block when it is determined that all the blocks in the row in which the block is completed are reconstructed.

Embodiment 3:

In the video decoding method, as shown in FIG. 9, when the reconstruction of the row is completed in step 102, a row reconstruction end message is sent to the entropy decoding module, and step 103 may be specifically:

Step 901: When the reconstruction module determines that all the blocks in the row of the block complete the reconstruction, the reconstruction module sends a line reconstruction end message to the entropy decoding module, which can still be referred to FIG.

Step 902: The entropy decoding module detects, according to the line reconstruction end message, whether the reconstruction module completes the reconstruction of the last row in the image frame. If yes, step 903 is performed. Otherwise, it waits for a certain time, and continues to execute 902 after waiting. Specifically, after obtaining the row reestablishment end message, the entropy decoding module may determine, according to the idle state information, whether all the rebuilding modules are in an idle state, and if yes, indicating that the rebuilding module has completed reconstruction of the last row in the image frame.

Step 903: The entropy decoding module performs entropy decoding on the code stream corresponding to the next input image frame, and further triggers reconstruction on the next input image frame.

Correspondingly, the entropy decoding module 201 in the video decoder of this embodiment specifically includes the structure shown in FIG. 10:

The decoding module 1001 is configured to perform entropy decoding on the code stream corresponding to the input image frame, and, according to the idle state information of the reconstruction module 202, distribute the entropy decoded row data to the idle reconstruction module 202 one by one in a predetermined order;

The detecting module 1002 is configured to: when the reconstruction module 202 determines that all the blocks in the row of the block complete the reconstruction, receive the line reconstruction end message from the reconstruction module 202, and detect whether the reconstruction module 202 completes the last line in the image frame according to the line reconstruction end message. The reconstruction, if so, triggers the decoding module 1001 to entropy decode the code stream corresponding to the next input image frame.

The following points need to be further explained:

1. The above embodiments can be applied to standards such as H.264, AVS, MPEG2, MPEG4 or HEVC.

2. In each of the above embodiments, the block may be a macroblock or a maximum coding unit (Largest Coding). Unit, LCU), etc.

The above content is a further detailed description of the present application in conjunction with the specific embodiments, and the specific implementation of the present application is not limited to the description. For the ordinary person skilled in the art to which the present invention pertains, a number of simple deductions or substitutions may be made without departing from the spirit of the present application.

Claims

A video decoding method, comprising:

The entropy decoding module performs entropy decoding on the code stream corresponding to the input image frame, and the entropy decoding module distributes the entropy decoded row data to the idle reconstruction module one by one according to the idle state information of the reconstruction module;

The reconstruction module reconstructs the row in the image frame according to the received row data, and sends a row reconstruction end message to the entropy decoding module when the row reconstruction is completed;

The entropy decoding module detects whether the reconstruction module completes reconstruction of the last row in the image frame according to the row reconstruction end message, and if so, entropy decodes the code stream corresponding to the next input image frame.
The video decoding method according to claim 1, wherein the reconstructing module reconstructs the rows in the image frame according to the received row data, specifically:

The reconstruction module extracts block data from the row data;

The reconstruction module detects whether the reconstruction of the block has a reconstruction condition.

If so, the reconstruction module reconstructs the blocks in the row according to the block data;

After the reconstruction of the current block is completed, the reconstruction module updates the reconstruction state information of the current block as a basis for detecting whether the reconstruction of the unprocessed block has a reconstruction condition.
The video decoding method according to claim 2, further comprising: after the reconstructing module updates the reconstruction state information of the block,

The reconstruction module determines whether all blocks in the row of the block have been reconstructed.

If so, the reconstruction module filters and interpolates the row of the block.
The video decoding method according to claim 2 or 3, wherein the reconstruction condition refers to an inter prediction condition or an intra prediction condition.
The video decoding method according to any one of claims 1 to 4, wherein the block is a macroblock or a maximum coding unit, and the video decoding method is applied to the H.264, AVS, MPEG2, MPEG4 or HEVC standards.
A video decoder, comprising: an entropy decoding module and a reconstruction module, wherein:

An entropy decoding module, configured to perform entropy decoding on a code stream corresponding to the input image frame, and, according to the idle state information of the reconstruction module, distribute the entropy decoded row data to the idle reconstruction module one by one in a predetermined order;

a reconstruction module, configured to reconstruct a row in the image frame according to the received row data, and send a row reconstruction end message to the entropy decoding module to trigger the entropy decoding module to detect whether the last line of the image frame is completed when the row reconstruction is completed. The reconstruction is performed corresponding to the entropy decoding of the code stream corresponding to the next input image frame.
The video decoder of claim 6 wherein the reconstruction module comprises:

a block reconstruction module, configured to extract block data from the row data, and detect whether the reconstruction of the block has a reconstruction condition, and if yes, reconstruct the block in the row according to the block data;

And a progress update module, configured to: after the block reconstruction module completes the reconstruction of the current block, update the reconstruction state information of the current block as a basis for the block reconstruction module to detect whether the reconstruction of the unprocessed block has a reconstruction condition.
The video decoder of claim 7, wherein the reconstruction module further comprises:

a judging module, configured to determine whether all blocks in the row of the block are completely reconstructed;

The filtering and interpolation module is configured to filter and interpolate the row of the block when it is determined that all the blocks in the row of the block are reconstructed.
The video decoder according to claim 7 or 8, wherein the reconstruction condition refers to an inter prediction condition or an intra prediction condition.
A video decoder according to any one of claims 6 to 9, wherein the block is a macroblock or a maximum coding unit, and the video decoder is applied to the H.264, AVS, MPEG2, MPEG4 or HEVC standards.