The Research On Rate Control Techniques Based On Statistical Model In H.264

Rate control is an important technique of video compression encoding. Its role is to control the quality of compressed video and the output stream to satisfy certain requirements of the transmission or storage. H.264 is the newest video coding standard developed jointly by ITU-T and ISO/IEC, and R-D optimization is introduced in H.264 for selecting the appropriate mode of motion estimation and macro-block. However, this could cause a well known chicken and egg dilemma. Therefore, a lot of conventional rate control algorithms are not applicable in H.264. Although the JVT-H017 algorithm can solve the dilemma, the R-D model is not precise enough and do not consider the impact of scene change, so it requires further research and improvement.Based on the analysis of the rate control algorithms of JVT-H017, as well as the statistical distribution of the DCT coefficients, this thesis has made deep study the rate control. At first, Cauchy model has in-depth analysis and research and simplified Cauchy R-D model in H.264 , and then the algorithm was described in details. Secondly, the thesis has study on the complexity of the image prediction method for further improvement based on Cauchy distribution R-D model. As the scene changes or a severe case of motion, the use of MAD prediction complexity of the image will fail, this thesis introduces the PSNR factor and MAD jointly to predict the image complexity method, and also the algorithms are given by simulations. The experimental results show that the new algorithm obtains precise and stable rate control compared with the MADr atio algorithm. Finally, take into account the existing H.264 rate control algorithm can not adapt the scene change problem, a simple and effective scene change detection algorithm is proposed based on Cauchy R-D model. The detection threshold was obtained by using statistical method, and a lot of scene change video sequences were tested. And also the experimental results proved that the new algorithm has higher video quality and more stable output stream compared with JVT-H017, and its additional computational complexity is not high and has practical application values.