Rate Control Based On Support Vector Machine In HEVC

With the popularity of high definition video and the development of communication technology, watching high-definition video in wireless channel or network of business has been paying more attention. Video is the main application of multimedia technology, the transmission of video in limited channel environment will need to use one of the core technology of HEVC:rate control. The current rate control algorithm cannot control the bitrate properly depend on the video characteristics, it may cause the overflow of buffer even intuitive visual quality decline.In order to overcome the problem above, this paper studies using the classification characteristics of support vector machine (SVM) to solve the bit allocation problem with different texture characteristics of video itself. This article first introduces the HEVC coding framework and classical bit rate control JCTVC-K0103, K0103is mainly introduced the whole control process; then this article introduces the principle of support vector machine and how the two classification model transmit into multiple classification model, through the experimental comparison and analysis of the classification model and directed acyclic graph is select as the most suitable model for this article. On this basis this article classified the output bits as a video coding standard, and the video classification will directly connect to the texture characteristics of the video itself. Then, using the trained model to classify the current video frame, and further to determine the number of bits of current frame based on texture features for bit allocation or video frame. For R-lambda update parameters of the model can’t update itselfs according to the scene to switch problem, in this paper, the application of support vector machine classifier to detect scene frame, and on the basis of the current frame texture properties to update the parameters of R-lambda model.The experimental results show that the proposed rate control algorithm has made a progress against JCTVC-KO103on BD-rate, and the bit error rate is controlled in a certain range.