Research And Improvement Of Rate Control In H.264/AVC Video Coding Standard

With the fast development of mobile communication and Internet technology, video conference, stream media communication and consumer electronics have been more and more influencing in common people’s life. In-depth studies have been made by the academic world and the industry. Through improved syntax structure, prediction algorithm and data transformation method, H.264/AVC, a new generation video compression algorithm, possesses fine coding efficiency, high definition, error-resilience performance and network access performanceRate control is one of the key technologies in video coding. With efficient rate control algorithm, the video coder can control the rate of output streams and maintain steady coding quality, while meeting the limitation of bandwidth and buffer capacity. Rate control algorithm in H.264/AVC has been more complex than in other video coding standard due to the wide application of Rate Distortion Optimization (RDO) technology. At first in this paper we make an introduction to the basic theory and system architecture of rate control. With the in-depth theoretical analysis and experimental results, we point out the problems of the current algorithm, including the bit allocation error in frame layer, the invalidation of scene change detection and treatment.Then a new parameter, called PLDR (Pixel Luma Difference Ratio), is proposed to describe the picture dynamic complexity comparing to the last few coded frames. It is calculated before the coding of each frame to effectively detect the movement of frames. This new parameter is also used in the abrupt scene change detection and the following treatment. A twice-pass coding method is introduced to deal with the particular scene change frames and guarantee the quality of the following frames. By introducing the PLDR and PSNR ratio, we have also improved the frame layer bit allocation model in H.264/AVC rate control scheme.An improved Rate Control Algorithm for H.264/AVC is proposed at the end of this paper. Conclusions can be made from the experiments that the new algorithm is more effective and more efficient than the formal ones such as JVT-G012 algorithm, especially in frame layer bite allocation and abrupt scene change detection and treatment. Higher PSNR and more precise bit rate are available while the subjective assessment of reconstructed frames is also better.