VC-1解码器技术参考手册

需积分: 50 177 浏览量更新于2024-08-02 收藏 313KB PDF 举报

"微软的VC1编码标准是视频编码领域的一个重要规范，主要由微软公司制定，WMV（Windows Media Video）格式就是遵循这一标准的典型代表。VC1（Video Coding 1）标准旨在提高视频压缩效率，降低存储和传输视频数据所需的空间，广泛应用于高清视频内容的制作和分发。此技术文档是VC-1解码器软件的技术参考手册，版本为2.1，由ARM Ltd在2005年为微软公司编写，并提供给了SMPTE（电影与电视工程师学会）。" VC1编码标准是基于早期的WMV9编码格式发展而来的，它在保持高质量视频的同时，提供了更高效的压缩比。该标准被广泛应用于DVD、蓝光光盘以及流媒体服务中。在技术实现上，VC1采用了先进的视频编码技术，包括宏块预测、运动补偿、熵编码和去块效应滤波等，以实现对视频数据的有效压缩。文档的主要内容分为以下几个部分： 1. 关于本文档：这部分介绍了文档的变更控制、当前状态以及变更历史，为读者提供了文档更新的背景信息。 2. 范围：明确了文档所涵盖的范围，可能包括VC1解码器软件的具体功能和应用领域。 3. 引言：对VC1编码标准和解码器的基本概念进行简要介绍，为后续章节做铺垫。 4. 项目结构：详细描述了解码器软件的目录和模块结构，以及设计决策，包括类型抽象、内存分配策略等。这些设计决策有助于优化软件性能和兼容性。 - 目录和模块：解码器软件的组织结构，可能包括不同的源代码文件和库，用于处理视频帧的解码、宏块处理、熵解码等不同任务。 - 设计决策： - 类型抽象：通过封装数据类型，提升代码的可读性和可维护性。 - 内存分配：详细说明了内存管理策略，如内存池或动态分配，以提高解码效率并减少内存碎片。 4.2.3后面的描述由于资料不全，无法给出具体细节，但可以推测会涉及解码过程中的其他关键设计决策和技术细节，如错误检测和恢复、多线程支持等。 VC1编码标准是视频编码技术的重要组成部分，其解码器软件的技术参考手册为开发者提供了实现和优化解码器的详细指南，确保了视频内容在各种平台上的流畅播放和高效处理。这份文档对于深入理解视频编码原理、优化视频处理系统或开发相关应用具有极高的价值。

VC-9 Decoder Software Technical Reference Manual

version 2.1 Page 16 of 137

Figure 4 - decoder example

5.4 Details of decoder operation

The following sections describe, in order, the stages involved in decoding a picture, from

bitstream input to YUV pixel output.

5.4.1 Bitstream unpack

The sequence, picture and macroblock layers of the bitstream are unpacked according to

sections ‘Bitstream syntax and semantics’, ‘Progressive bitstream syntax and semantics’ and

‘Interlace syntax and semantics’ of the standard [1]

, and the data they represent stored in the

decoder’s state structure.

5.4.2 Prediction

The CBP, MV, DC and AC coefficients are predicted for the current block based upon the

previous blocks, as described in sections ‘Coded Block Pattern’, ’Motion Vector Predictors’, ‘DC

Predictor’ and ‘AC Prediction’ of the standard [1]

respectively.

5.4.3 DC coefficient decode

If the picture is type I, or the block is intra type, the quantised DC coefficient is extracted from the

block layer, added to the predicted DC coefficient, and stored in the first element of the transform

coefficient array. The predicted DC coefficient is added at this stage.

5.4.4 AC coefficient reconstruction

If the CBP indicates AC coefficients are present in this block, the algorithm described in the

figure named ‘Coefficient decode pseudo-code’ of the standard [1]

is used to obtain each run and

level pair. These are used to reconstruct the AC coefficients of the transform coefficient array

according to the algorithm described in the figure named ‘Run-level decode pseudo-code’ of the

standard [1]

. These are stored in a temporary array.

Configuration = Initial Decoder Configuration;

vc1DEC_DecoderRequirements(&Size, &Configuration, &Bitstream);

State = malloc(Size);

vc1DEC_DecoderInitialise(State, &Configuration);

vc1DEC_DecodeSequence(State, &Bitstream);

while ( Bitstream not exhausted )

{

vc1DEC_DecodePicture(State, &Bitstream, &Picture);

/* display or write picture to disk */

}

vc1DEC_DecodeFlush(State, &Picture);

/* display or write picture to disk */

VC-9 Decoder Software Technical Reference Manual

version 2.1 Page 17 of 137

5.4.5 Zigzag scan

If the CBP indicated AC coefficients are present in this block, the temporary transform coefficient

array is rearranged using a zigzag scan array, and the result put into the transform coefficient

array. The results are combined with predicted AC coefficients, if AC prediction is enabled.

5.4.6 Dequantise

The DC coefficient is dequantised according to the algorithm described in section ‘DC Inverse-

quantization’ of the standard [1]

If the CBP indicated AC coefficients are present in this block, they are dequantised according to

the algorithm described in section ‘Inverse AC Coefficient Quantization’ of the standard [1]

. The

transform coefficient array is dequantised in place.

5.4.7 Inverse transform

The inverse transform operation is performed according to the specification in the annex ‘Inverse

Transform Specification’ of the standard [1]

. The operation is performed on the transform

coefficient array, in place.

Unclamped results from the macroblock are stored for use in overlap smoothing.

5.4.8 Block prediction

If the picture is type P, a macroblock is obtained, interpolated if necessary, from the reference

picture, and combined with the resulting macroblock from the inverse transform.

If the picture is type B, one or two macroblocks are obtained, interpolated if necessary, from one

or two reference pictures, averaged and combined with the resulting macroblock from the inverse

transform.

The macroblock is clamped and copied into the key frame buffer in the decoder’s state structure,

for future prediction use.

5.4.9 Post process

If enabled, overlap smoothing is performed as described in section ‘Overlapped Transform’ of the

standard [1]. Data from previous macroblocks required to perform the operation is stored in the

decoder state. The image is also deblocked, if the bitstream requires it.

5.4.10 Display

The key frame buffer is copied to the display buffer, which is returned to the application. Range

reduction and resolution scaling is performed at the same time, if required by the bitstream.

5.5 Decoder API functions

This section describes the API functions used by the decoder. For more detailed internal

structure and function documentation, see section 8, VC-1 Data Structure Documentation, and

section 9, VC-1 File Documentation.

VC-9 Decoder Software Technical Reference Manual

version 2.1 Page 18 of 137

5.5.1 Decoding

5.5.1.1 vc1DEC_DecoderRequirements()

This function is called once per bitstream. It will examine the bitstream and calculate the amount

of memory that the application must allocate for the decoder to decode the bitstream.

5.5.1.2 vc1DEC_DecoderInitialise()

This function initialises the area of memory allocated by the application for use by the decoder.

5.5.1.3 vc1DEC_DecodeFrame()

This function is called once for each picture to be decoded. It will decode a whole or part of a

picture, depending on whether the picture is split into slices or fields. It will write the resulting

data into the decoder’s picture pipeline if a complete picture is available, otherwise further calls

to vc1DEC_DecodeSlice() or vc1DEC_DecodeField() will be required. A picture will be copied to

the application’s picture buffer if the pipeline has been filled, which will be indicated by the

function’s return value.

5.5.1.4 vc1DEC_DecodeFlush()

This function is called once per bitstream, at the end of a decoding session. It flushes the

decoder’s picture pipeline, and writes the picture to the application’s picture buffer. No bitstream

is consumed, and no decoding operations are performed.

5.5.1.5 vc1DEC_DecodeSequence()

This function is called when a sequence layer BDU is received. It decodes the layer, and puts the

information it contains into the sequence parameters structure, which may be accessed by the

application.

5.5.1.6 vc1DEC_DecodeEntryPoint()

This function is called when an entry point layer BDU is received. It decodes the layer, and puts

the information it contains into the sequence parameters structure, which may be accessed by

the application.

5.5.1.7 vc1DEC_DecodeSlice()

This function is called when a slice layer BDU is received. It decodes the layer, and any image

data it contains. The decoder’s internal picture buffer is updated with more image data. If the

slice completes a picture, the picture is returned to the application. Otherwise, further calls to

vc1DEC_DecodeField() or vc1DEC_DecodeSlice() may be required. This is indicated by the

function’s return value.

5.5.1.8 vc1DEC_DecodeField()

This function is called when a field BDU is received. It decodes the image data in the field, and

updates the decoder’s internal picture buffer. If the field completes a picture, the picture is

returned to the application. Otherwise, further calls to vc1DEC_DecodeField() or

vc1DEC_DecodeSlice() may be required. This is indicated by the function’s return value.

5.5.1.9 vc1DEC_UpdateBuffers()

This function is called to update the Hypothetical Reference Decoder model within the decoder at

the end of decoding a frame. The function will return vc1_ResultHrdUnderflow or

vc1_ResultHrdOverflow if any of the HRD buffers underflow or overflow respectively.

VC-9 Decoder Software Technical Reference Manual

version 2.1 Page 19 of 137

5.5.2 File handling

The application must provide bitstream reading and YUV buffer output. See the reference code

for an example of how to update the decoder library bitstream information.

6 DECODER APPLICATION

6.1 Access to bitstream information

The decoder library parses many informational items from the bitstream which are potentially of

interest to the calling application, but are not used within the decoder library itself. The library

stores this information within the allocated state information. If the application wishes to read

particular values, it can do so via the allocated state structure:

• pState-> sSeqParams provides the sequence layer and entry point layer information

• pState-> sPicture provides information on each output picture.

6.2 Option processing

The decoder requires a single command line parameter when run. This parameter is the name of

a simple text file that defines all the particular options used by that run of the program. The

option file can define options for use by the decoder only, or other associated tools – simple

conditional inclusion allows tool-specific options to be parsed only when appropriate.

This code is not part of the operation defined by the standard, so the modules do not use the

vc1

prefix.

6.2.1 Option file format

Option lines are name and value pairs, with the name separated from the value by a colon (‘

’).

For options controlling arrays, an index in square brackets, such as [0], can be added after the

option name. Comments are allowed; any text following a ‘

‘ or ‘

;

’ character will be ignored

(unless it is recognised by the conditional processing described below). Case is ignored

throughout. Whitespace is also ignored throughout (except for the values for string-type options,

which have leading and trailing whitespace removed).

Conditional processing recognises only

#if

and

#endif

directives. These cannot be nested.

The condition consists of the tool name, or a list of tool names separated by the C logical-OR

operator (‘

’).

Here is a simple option file to demonstrate the syntax.

# Options for SMPTE VC-1 reference decoder.

# Comments in this file can start with # or ; at any point in the line

; Blank lines are allowed.

#if decoder

DebugMask : 0x80000000 # See debug.h for bits to set/clear

#endif

Picture width : 176

Picture height : 144

剩余136页未读，继续阅读

imagany

粉丝: 0
资源: 4

VC-1解码器技术参考手册

SMPTE 421M标准：微软VC-1视频压缩协议详解

VC-1视频编码在DM6467上的实现与优化研究

VC-1压缩视频位流格式与解码过程标准文档

微软的视频编码标准VC1（wmv9）标准文档

vc1编码器，可以编码vc1格式的视频

VC9微软编码标准草案

VC1 技术文档

微软的VC-1视频编码格式介绍

WMVHD视频编码MicrosoftExpressionEncoderV3.0.1332绿色汉化版

SMPTE公布的VC-1标准文档

最新资源