Abstract
As content based video indexing and retrieval has its foundations in the prime video structures, such as a shot or a scene, the algorithms for video partitioning have become crucial in contemporary development of digital video technology. Conventional algorithms for video partitioning mainly focus on the analysis of compressed video features, since the information relevant to the partitioning process can be extracted directly from the MPEG compressed stream and used for the detection of shot boundaries. However, most of the proposed algorithms do not show real time capabilities that are essential for video applications. This paper introduces a real time algorithm for cut detection. It analyses the statistics of the features extracted from the MPEG compressed stream, such as the macroblock type, and extends the same metrics to algorithms for gradual change detection. Our analysis led to a fast and robust algorithm for cut detection. Future research will be directed towards the use of the same concept for improving the real-time gradual change detection algorithms. Results of computer simulations are reported.