Research And Application Of Video Compression Based On ROI

In the video communication, video compression and video transmission are two most important parts. Video communication part of ITF project has also met the problem of Qos. This paper is mainly about video compression and aim at improving video quality through ROI based video compression technology.Research work include following aspects: 1. Segmentation of ROI from video; 2. Strategy of compress ROI; 3. Design of video transmission part and the optimization of video compression.First of all, this paper propose a solution of ROI segmentation based on analyzing the characteristics of human vision, including physical ones and psychological ones, and making the conclusion that the ROI in the video are regions where image grads is high. And also the human face region. This paper proposes segmentation algorithm based on image grads, and human face detection algorithm based on distance transformation. This human face detection algorithm makes progress on skin-color segmentation algorithm, and improving the speed of locating algorithm. Moreover, algorithm of this paper is also efficient to multiple face condition.Secondly, about the strategy of ROI compression, an algorithm of video preprocesses is put forward. It use variable variance Gaussian filter to blur the image of non-ROI. And the ROI can get more bit to use. Consequently it can improve ROI quality. About compression control, a strategy that different quantize parameter to ROI and non-ROI is proposed. This strategy adjust quantize parameter according to network status. It can make ROI encoded with high quality first.At last, a design solution of video transmission system is proposed. And then this paper makes use of layered motion estimation in non-ROI and Intel multimedia instruction to optimize video compress, improving the capability of real time.