The Research On Video Transcoding Based On Region Of Interest

With the development of wireless network, more and more consumers are becoming increasingly enthusiastic in enjoying video via mobile deviecs. There are two characteristics in enjoying these videos with mobile terminals. One is the low-bandwidth of wireless environment, it can still hardly meet the high-quality video requirements; the other critical constraint is the limited display size and low resolution of mobile devices, the existing video programs can not be provided to these terminals directly. Video transcoding based on region of interest is an affective solution.Ground on the summary of the existing techniques for video transcoding while consider the characteristics of mobile devices, this thesis does a research on spatial transcoding algorithm and frame skipping trancoding algorithm based the visual attention model. The main work of this thesis is summarized as follows:Firstly, this paper analyses the main visual attention modeling methods proposed in recent years. The visual attention model which simulates the viewers' visual response and perception by integration of the static and dynamic visual features of video content is an effective representation of video contents.Secondly, a fast transcoding scheme based on region of interest is proposed for mobile video terminals in this paper. According to the target displaying size, the region of interest is automatically detected from the input H.264 video stream in the video server, and then generates the video stream, which is suitable for viewing on mobile video terminals, at the proxy server by video transcoding. Moreover, a fast mode decision algorithm is proposed for this scheme. Experimental results demonstrate that the proposed approach can not only reduce the bandwidth occupation, but also obtain excellent video quality with a low computational complexity.Thirdly, a frame skipping transcoding algorithm based on visual attention is proposed. After the new frame rate and target bandwidth are determined, the sliding window length can be obtained, the proposed method considers the compressed bitsream information to measure whether a frame should be skipped or not by the visual attention in the sliding window, then the key frame can be encoded by the rear encoder in the transcoder as a new transcoded bitstream to fit the new channel bitrate. Experiment result shows that our algorithm can choose more significant frames to avoid the jerky effect and achieve better visual quality.