Research On Adaptive Quantization Technique For Video Coding

Along the improvements of processor capability both at the sending and receiving sides, and the broadening of network bandwidth, it is noted that the multimedia communication is coming to be the key component of next-generation communication system. As the inevitable part of the multimedia communication system, the video communication and video coding are more and more important and necessary. Co-sponsored under National Science Fund of China– Research on general Rate-Distortion Model on Wireless Channel (Grant No.: 60372067), University Cooperation Project of Huawei Technology Corporation– Research on the AVS video standard proposals, Youth Chenguang Program of Wuhan– Research on Joint Power-Rate-Distortion Model for Wireless Multimedia Communication (Grant No.: 20045006071-18) and High-Definition TV based subgroup of Audio Video Coding Standard Workgroup of China (AVS1-P2 ZQ Profile), we do the research on the video coding and participate into the AVS workgroup actively including proposing the technical proposals and pushing the standard advancing domestically.In this thesis, at first, the basic theory of hybrid video framework and the development of mainly video CODEC standards are introduced. The advantages of AVS are presented from the comparison with H.264/MPEG-4 AVC, including stream syntax, decoding processing and function of modules. Focusing on the HDTV application trend, a new macroblock based AVS decoder is designed. For the decoder part, the macroblock-based decoder (AVS Golden C decoder) can finish the stream parsing and decoding processing. The third part of the paper will introduce the framework and realization of AVS Golden C decoder.Based on the introduction on the video standards, technology and AVS Golden C decoder design, in the forth part, the research and development of adaptive scalar quantization on AVS platform are presented. On optimizing quantization offset, the relationship of bit rate and distortion was researched, and an adaptive quantization algorithm is presented. The algorithm was adopted by AVS Workgroup, and in this field we also published a high level paper. On analysis of structure of quantization matrix, focusing on the characteristic of AC parameters in B frame, an adaptive quantization method is proposed to save the bits without distortion. A patent was applied for this algorithm.