| Video, including not only the content of static images, but also the information of motorial objects in scenes and the time movement in the external world, is a kind of important visual resources. It is an important part in multimedia field. At the same time, how to find what we interested in from huge databases has been noticed by many researchers. Under these circumstances, we discussed the technique of content-based video retrieval. Firstly, we processed hierarchy structure partition to the video, and analyzed the necessity of events as the retrieval units from the point of rogatory granularity in video retrieval; Secondly, from the video editing, we proposed a new technique that extracted the events based on the representational shots in order to low the computational complexity and prepare for the last work. Lastly, in the clip-based retrieval, we use the above-mentioned events as our matching units. Complete matching based on Kuhn-Munkres algorithm was then adopted to obtain the final similar video events. In our paper, we used the films as our tested videos and the results showed that they were accorded with the fact. |
