Content-based Video Retrieval System

With the rapid development of multimedia, digital television, network technique and wide application of science and technology, digital video made great strides, generating a new type of media carrying enormous volume of information. How to organize, manage and use the digital information have become an imperative and a heated subject in multimedia field.Shot is the basic unit that composes video. The accuracy of shot segmentation has direct influence on post-treatment and retrieval effect. So this thesis works on the shot accuracy in content-based video retrieval and time of retrieval to figure out a real-time and self-adaptive method of shot detection. The method can detect abrupt change shot effectively by calculating neighbor frame difference, the gradual change shot by non-neighbor frame difference, which have attained good comprehensive detection effect.In addition, this thesis developed a content-based video detection system. Its key part is to use clustering method based on accurate shot segmentation to realize precise generation of scene and story unit. This system applies delamination retrieval, which conquered the aimlessness of former retrieval technique and put the object on the scene layer, largely shortening detection time and increasing the efficiency.This thesis consists of five parts. The first introduction part talks about the current situation and the main job on content-based retrieval research. The second section talks about relevant knowledge of the histogram technology.The third part mainly expounds the related basic theory of video analysis and retrieval. The fourth part introduces the real-time and self-adaptive shot detection system based on neighbor frame difference and non-neighbor frame difference. By introducing partitioned histogram to reflect the spatial qualities of video image, it improved the effect of shot segmentation. The fifth part discusses a content-based video detection system. In the end it summarizes the research content in this thesis and the prospect for the work to come.