Categorizing video shots by utilizing SVM and wavelet

摘要

Shots classification plays an important role in well indexing, browsing and retrieving video content. By that, the large amount of video content can be efficiently indexed, and then, it can provide convenience for managing video. In this paper, edge features are firstly extracted by wavelet, which can not only reduce amount of shots data but also preserve the important structural properties of shots. And then, to reflect local properties of shots, ratio of edge pixels in each sub-window is calculated. After that, color moments are computed to reduce loss of global properties, which can assist edge features in well indexing shots. Finally, support vector machine (SVM), which has a very good performance on pattern recognition, is employed to classify shots. Experimental results demonstrate that this method can efficiently categorize video shots and satisfy the basic needs of shots classification.