Research On Video Transcoding Based On AVS In Transform Domain

Along with constantly updated video coding standard, diversity of video terminals, and heterogeneous network enviroment, video transcoding technology has become the core of multimedia application technologies. AVS is the second source coding standard with independent intellectual property in China, it has very broad prospects with high coding efficiency, low computational complexity and low patent costs. However, the further development of AVS is limited by the relative lack of program resources, transcoding from more abundant and mature MPEG-2 resource to AVS standard is required necessarily.Based on the mainstream video coding standards, the advanced transcoding tecnology is studied deeply in this paper. By contrasting and analysing the key technologies of transform, quantization, intra prediction, inter prediction, entropy coding and loop filter, and scope of various transcoding technologies, a new transcoding structure in transform domain from MPEG-2 to AVS is proposed. Intra and inter transcoding are introduced and researched in detail to solve a variety of problems, such as conversion from DCT to AT domain, decision of intra prediction mode, judgement of inter macroblock segmentation, conversion from B to P frame, and motion vector reuse. The theoretical analysis and design of core modules of transcoding system are completed in this paper, and the whole transcoding system from MEPG-2 to AVS is implemented.Experimental results through different test sequence show that, in case of image quality, the accuracy of macroblock mode prediction is reached to 64%, the overall transcoding time is reduced by 45%, the whole system saves 50% computational complexity comparing with traditional cascaded transcoding system, meeting the demand of real-time transcoding.