A Data-Mining Based Video Shot Classification Method

A novel classification method of video shot genre based on data-mining has been proposed. Shot boundary detection and key frames extraction are firstly performed. Then, some visual features such as color and motion are extracted for the key frame and shots. Furthermore, decision tree is applied to discover the rules between these features and shots genres from numerous training data. These rules are finally exploited to classify the new video shots. Experimental results show that, compared with the method based on SVM (Support Vector Machine), the proposed method can achieve higher detection accuracy and the rules obtained are easy to comprehend.