A novel framework of shot boundary detection for uncompressed videos

The automatic video shot detection is receiving a great impact with the advances in the digital video technology and ever increasing accessibility of computing results. In this paper we describe a framework for extracting shot detection by using the threshold values of diverse statistical features for raw video frames. Two different types of sports videos viz. soccer and basketball are used for assessment. The approach exploits correlation, maximum histogram difference and running average difference as the classifiers. The results are evaluated by selection of appropriate threshold of these features after training of framework. The winner take-all selection scheme is applied if correlation coefficient and histogram difference features are unable to identify the shot detection. Experimental results on divergent set of test videos reveal the effectiveness of this shot detection approach.