Research On Shot Boundary Detection And Video Hierarchical Organization

With the development of computer, communication and multimedia technologies, the amount of video data is enlarged at a explosive speed. How to orgnize, manage, express and retrieve these video data efficiently becomes a hot issue on video retrieval. This leads to the emergence of content-based video retrieval technology. Shot boundary detection and video hierarchical orgnization are discusseded in this paper.Firstly, features which used to detect shot boundary are discussed . Based on the analysis of existing techniques of cut detection, a new idea which uses support vector machine (SVM) to detect cut position is proposed. Since SVM has a good generaliazation ability under the condition of limited training samples and cut detection can be seen as a two classes problem in pattern recognition, a SVM classifier is trained for determining cut positions. The simulation results of different video sequences demonstrate that the proposed method is robust and has a good performance of the generaliazation.Secondly a new approach based on wavelet denoising for shot boundary detection is presented. The algorithm can detect not only cut but also gradual change. The experimental rusults indicate that the proposed approach is of good precision and generality.Finally, the subject makes an analysis on the idea of scheme for the hierarchical orgnization of video.In view of this idea, a new clustering method based on immune and clonal are used for video scenes clustering. The experiments show that the new clustering method has satisfactory performance. Based on this, a kind of hierarchical video content organization can be realized, which makes a good condition for further video processing.