Research On Video Coding Fast Algorithms And Complexity-Distortion Model Based On H.264

With the economic development and technological progress, the market has increasing demands of high-performance video service. In this case, the previous video coding standards can't meet the requirement. In 2003, the ITU-T VCEG and the ISO/IEC MPEG jointly developed a new generation video coding standard H.264/AVC, which has higher compression efficiency and improved network adaptation. Hence it will have a broad application in digital video communication and storage areas in the future.However, the high compression efficiency of H.264 is at the cost of high computational complexity, which has become the bottleneck in the realization of software and hardware, therefore it is very important to develop a low complexity video encoder. Furthermore, in the condition of software coding and decoding, the available processor resource will change with the different platforms or assignment quantities occupying the processor resource, which requests that the encoder has unfixed computational complexity. The handset drived by the battery also requires that the coding complexity can be controlled considering the limited power consumption. Currently,the research on fast coding algorithms and complexity- distortion for H.264 have been one of the hotspot issues in both industry and academia. The following creative studies are performed according to the two aspects:Firstly, for the time-consuming modules of inter-frame mode selection and motion estimation in H.264 reference encoder JM, an adaptive inter-frame mode decision algorithm and a multi-reference frames fast selection algorithm are proposed respectively in this paper. On the basis, local codes of JM are optimized and a low complexity video encoder JM_OPT is developed.Secondly, a complexity scalable motion estimation arithmetic aimed at JM_OPT is brought forward on the basis of analyzing the influences of complexity, distortion and encoding bit-rate of coding parameters in the motion estimation module synthetically. And taking account of the characteristics of integer transform, quant and zig-zag scan in H.264, a complexity scalable DCT arithmetic has been produced based on the DCT pruning thought. A complexity scalable video coding framework based on parameters has been produced for the JM_OPT by taking the two above algorithms into account synthetically, which can realize the complexity scalability by adjusting its complexity control parameters, thereby a good balance between the video quality and complexity can be formed.On this foundation, this paper proposes an optimized complexity distribution algorithm among frames and a complexity-rate combined control strategy according to the features of complexity control and rate control. And a complexity-distortion model is established, which can adjust the computational complexity adaptively according to the change of calculation capability or power consumption of terminal processor.Finally, simulations for the C-D model are carried out in the smart phone simulator on the thought of the calculation capability, memory and energy consumption of smart phone, and the results validate the feasibility of this model.