Smart video compression

点击翻译按钮获取中文摘要

This research is a collaborative effort between a research team at the University of Ottawa and a Canadian industrial partner, Thales Defence and Security Canada. Thales Canada is the prime contractor for the Land Command Support Systems (LCSS) providing long-term software design, development and integration support of the Canadian Armed Forces' core command and control system. It has developed a solution for video surveillance with a vehicle but in the field the video needs to be compressed to be sent back quickly to headquarters and to be stored over the long term. Standard video compression technology is not suitable because the required compression ratios cannot be achieved without losing important information. The proposed research project will develop new solutions with video summary techniques based on content analysis, user interests and long-term detection, tracking and learning. These techniques can achieve much higher compression ratios than general purpose video encoders because the overwhelming number of frames in a surveillance video record nothing but uninteresting background. The expertise of the University of Ottawa researchers in saliency models, content aware compression and tracking will enable the rapid development of software technology for content-aware video compression and summary based on user interactions. This software will be developed at the University of Ottawa in tight collaboration with Thales' Ottawa Office and delivered to Thales Canada at the end of this project.