Research On The Key Technology Of Content-Based Video Retrieval

With the rapid development and spread application of the computer technology, all sorts of multimedia data can be collected, produced and transmitted by manifold means. Among these data, video has the largest mount and greatest complexity. It becomes more and more difficult for people to manage and search video information efficiently because of its rapid increase. Therefore, how to organize and manage video data effectively becomes a subject to solve urgently. People can search and make good use of video while this problem was successfully resolved. Under the background, Content-based Video Retrieval (CBVR) becomes a hot research topic that researchers care about strongly and universally in recent years. This paper has done the following groundwork around several key technologies of this field.Firstly, the algorithms of shot boundary detection were researched in the paper. An advanced algorithm to double threshold algorithm was proposed via the analysis of abrupt, gradual change and flash. A slid window was applied in this improved algorithm to detect shot boundary using the local information around the shot transition. Several types video were segmented into shots adopting this advanced algorithm and double threshold algorithm in experiment. With the result, we can test the detection efficiency of the advanced algorithm in shot boundary detection.Secondly, the technology of key frames extraction was explored helpfully. The principle and characteristic of key frames extraction were studied, and the algorithms of this filed were briefly analyzed. Then two effective algorithms were put forward. One is based on several confirmations, the other is based on hierarchical clustering and graph center. At last, different types video shot were extracted key frames in the experiment while comparing to other algorithms.Experiment results can prove the extraction efficiency of these two methods in key frames extraction.Lastly, a universal retrieval system framework for Content-based video was designed according to design requirement of retrieval system. This system framework was built on the former job of this paper. In order to achieve better analysis and retrieval function, it is composed of video pretreatment subsystem and video query subsystem.