[codec] write H264 decoder (1) from scratch

Hello everyone , I meet you again , I'm your friend, Quan Jun .

1. background

Because of work , I have been in contact with codec for some time .AVC,HEVC, There are also some functions, large and small , I have been reading books and articles on the principle of codec . From introduction to a little understanding , It feels a little fragmented , Or incomplete , It doesn't seem to form a system . Some small functions , To know is to know , I don't know its meaning and function , Over time, you will gradually forget . Think about it , Maybe a lot of things , You still need to do it yourself , Will understand deeper and more thoroughly , Like Feynman's learning method , Can you say it , That means I understand , This one is the same , You can realize the function , That means you really understand the process and logic . So , During the Chinese new year this year , Suddenly came up with the idea of writing a decoder , And I couldn't hold it down as soon as I started , I've always wanted to move the keyboard and write . In fact, there are many open source and easy-to-use decoders on the market , image ffmpeg,x264 wait . Own this project , It should be a simple learning project , It is estimated that the effect and function of these famous projects will not be achieved in the end , But so what , The process and experience are also great , isn't it? ？ At the beginning, I wanted to write an encoder , After thinking about it, I soon gave up , My current idea is just to familiarize myself with the agreement , Not focusing on coding algorithms , by comparison , The knowledge needed to write a decoder is exactly what I need . This has become the cause of this series of articles , Write your own code , Write a summary . Although from “ zero ” Start , But the basic knowledge of codec still needs some reserves , I will introduce and explain the knowledge points involved in decoding in each chapter , But it's too fragmented , It won't be explained one by one . If the knowledge point is too big , You may write a separate article to summarize .

2. plan

2.1 Writing plan

Follow the steps of a decoder , I'm going to divide it into these parts for the time being

• NALU： Format , Start code , EBSP RBSP SODB
• Columbus entropy coding of entropy coding
• SPS： Analyze the meaning of each member in the agreement
• PPS： Analyze the meaning of each member in the agreement
• SEI： Analyze the meaning of each member in the agreement
• SLICE Header： analysis Header
• SLICE data： analysis slice data
• MarcoBlock： How to restore a complete macroblock YUV data
• macroblock_layer
• mb_pred
• Intra predicted part The prediction mode of
• CAVLC
• residual Residual data acquisition
• Intra predicted part Forecast data for and Image data recovery
• deblocking
• …… undetermined

2.2 Project plan

• development environment C Language
• Include features ： H264 It is divided 7 individual profile Of , Every profile Is a collection of different functions . See... For specific differences wiki Encyclopedias plan ： The first stage is the most basic one baseline The decoder of （ nothing B frame , Only CAVLC, Only line by line is supported Progressive） The second stage is to do main or Extend The decoder of （ Yes B frame , Yes CABAC, Support Progressive and interlace）

The follow-up will not be supported . Decoding does not display , from h264 File solution YUV File can .

• Time arrangement ： expect 6 Months to finish it baseline (8 month 30), The second stage will be arranged later .

3. Reference resources

3.1 Reference material

• H264 Official White Paper

3.2 Reference Engineering

• ffmepg Mirror
• x264 – mirror
• bitstream

3.3 Reference blog

1. Implement one from zero H.264 Stream resolver Supporting works
2. Do it yourself H.264 decoder -ZigZagSin Supporting works
3. H264 Video decoder C++ Project description Supporting works

4. Tool preparation

4.1 Bitstream analysis tool

• vega： I've used the most professional , Analyze the best software . However, it can only be used by units. Ha ha , Visual inspection should be expensive .

• H264 Visa : I am using 1.15 edition , Out-of-service vega Used when . The parameter analysis is comprehensive , It is very helpful to judge whether your parsing is correct in the early stage .

It is said that there is an upgraded version later H.264 CodecVisa, It doesn't work for the time being .

• elecard I tried it out , I don't think it's very easy to use , It is better to be beautiful and easy to use than vega, Wheel parameters are not as complete as H264 Visa. Not much use . Here is a record of ,elecard There are many kinds of software ,elecard stream analyzer, elecard stream eye,Elecard StreamEye Studio wait . Actually Elecard StreamEye Studio contains stream analyzer, Don't get confused by a bunch of names . among streamEye It has a comprehensive function .

4.2 YUV View tools

• 7yuv

4.3 player

• potplayer： see h264 stream

4.4 some YUV data source

• YUV Sequences (asu.edu)
• Xiph.org :: Derf’s Test Media Collection

Publisher ： Full stack programmer stack length , Reprint please indicate the source ：https://javaforall.cn/150705.html Link to the original text ：https://javaforall.cn