The Study On MPEG-2 To H.264 Video Transcoding Algorithm

Currently, MPEG-2 is video coding standard with widely application and accumulate much resource. With the continuous development of multimedia technology, video coding technology has developed rapidly. H.264 is the recently published new coding standard can achieve considerably higher coding efficiency than previous standards, and can meet different application requirements from the field of low bit- rate video (such as mobile phone) to the field of high bit-rate video (such as the high definition television) due to the flexibility encoding. Therefore, the research of fast transcoding algorithm for MPEG-2 to H.264 has very important significance of application.Firstly, the video transcoding technology research background and current status of researeh is introduced in this thesis,and the concept of digital video encoding, MPEG-2 coding standard H.264 coding standard are briefly description. Through a comparative analysis MPEG-2 coding standard H.264 coding standard, the similarities and differences beteen the two standards are sumed up, which supplies knowledge reserves with the research of MPEG-2 to H.264 algorithm. Secondly, in order to use the minimum complexity to achieve the best performance, we use MPEG-2 stream in the CMT, coded block pattern (CBP) and the DC coefficient of macroblock adaptive selection of H.264 coding mode. The results show that the algorithm significantly reduce the transcoding time while maintaining the video quality (an average reduction of 69.60 %.), compared to the fully encoding method.Thirdly, based on the hypothesis that MB coding mode decisions in H.264 video have a correlation with the distribution of the motion compensated residual in MPEG-2 video, we proposed a alogorithm .We use machine learning tools to exploit the correlation and construct decision trees to classify the incoming MPEG-2 MBs into one of the several coding modes in H.264. The proposed approach reduces the H.264 MB mode computation process into a decision tree lookup with very low complexity. Experimental results shows that the proposed alogorithm significantly reduce the transcoding time while maintaining the coding efficiency.Finally, we proposed a color video quality assessment based on edge-color distortion. The esponse functions of HVS to the edge distortion model and color distortion model are given by analyzing relation between predicted value of the two models and the subjective quality estimation. Using multiple linear regression analysis, the functions are integrated into an equation as the color video quality assessment that is used to evaluate video quality after transcoding.