Improving cut detection in MPEG videos by GOP-oriented frame difference normalization

The detection of abrupt shot changes ("cuts") in videos is a basic step in video content analysis. Many cut detection algorithms based on histogram differences have been proposed in the literature, and it has been shown for the MPEG (moving picture experts group) domain that using small sub-images, namely approximated DC-frames, is sufficient to achieve very good detection results. In this paper, the characteristics of histogram based difference measurements of MPEG (DC-) frames are analyzed and an effective technique is presented to enhance the performance of cut detection algorithms, called "GOP-oriented frame difference normalization" (GOP: group of pictures). Experimental results for the MPEG-7 video test set is presented to demonstrate the benefits of our proposal. Furthermore, the proposed method is not limited to a particular algorithm but it is applicable to an entire class of cut detection algorithms.