The Analyse Of Entropy Coding In AVS Video Encoder

Video coding technology is used in many fields such as DVD, HDTV, visible telephone, digital video meeting, and video surveillance and so on. In order not to pay a lot of patent money to international video groups, an organization named as"AVS standard working group"is developed in our courtry and brought forward Audio Video Coding Standard (AVS). This standard has many good characteristics, including good performance, low complexity algorithm, low implementation cost, simplified patent warranty and so on. We have reason to forecast that AVS can find many application fields. Entropy coding is in charge of compressing all the information to be transferred and is very important. The optimization of the entropy coding can improve the performance of the entire video encoder.As the research of entropy coding in AVS video encoder, the advantages and shortcomings of exp-glomb coding are analysed, the principle and the flow of AVS block transform coefficients are introduced, entire design scheme of AVS encoder is brought forward based on macroblock line parallel process technology. Each module and their functions are introduced in detail. The principle and working flow of macroblock line parallel process technology are analyzed as well as Data flux to be transferred. According to the hardware structure of DSP chip, entropy coding is optimized by intrinsics. The dimension of three dimension table is reduced. The initialization of exp-glomb is modified and block transform coefficient coding steps is simplified. Running time of entropy coding is shortened by these methods."Startingcode emulation"event is introduced which may happen when video encoder produces the bitstream to the file. The occurrence reason of pseudo startingcode and its affects to the decoded pictures are analysed. In view of this problem, a solution based on glomb code inversion is proposed. In the mean time, a new method of detecting pseudo starting code is put forward. This method modifes the structure of entropy coding so it avoids a lot of conditional judgements which take up plenty of DSP cycles, and thus the number of detection is reduced obviously. The test results show that after optimization entropy coding accords with AVS standard completely and the bitstream is the same as that before optimization.