Shot detection using a syntactic approach to the frame structure

Shot boundary detection is the first task of automatic video indexing. It is well known that all subsequent processes, such as video segmentation, video skimming [ 1 ], key frame extraction and video indexing, start from an accurate shot boundary detection. In the proposed paper the authors are investigating the possibility of using textual string extracted from the frame images, according to Chang and Fu [2], for shots identification. The results are comparable with other well known methods in cutting scene but they are more promising in detection of slow change for example in tilt, dolly, and pan.