Skip to Main Content
As the Internet and multimedia technology develops, the content security of the multimedia has become more and more important. To distinguish various contents in the multimedia, we present an approach for automatic video classification based on combination of MPEG-7 descriptors and second-prediction strategy. In this paper, color, texture, shape and motion descriptors are extracted from five different genres of videos and combined as a whole feature. Then we put the feature into the SVM classifier to be trained. We choose the 1-1 method for SVM multi-class classification, and use the second-prediction strategy to improve the accuracy of video classification. Finally, we test our approach on a broad range of video data and achieve an overall classification accuracy of 98.80%.