A Novel Shot Detection Algorithm Based on RS in Compressed-Domain

We propose a novel shot detection algorithm based on RS in Compressed-Domain. Firstly, the Algorithm extracts I frames from Compressed-Domain data sequence. We constructs information system with the difference between two adjacent I frames in column and attributes sets witch extracted from decompressed I frames in row. Then the established information system is normalized and discredited. Secondly, we reduce information system by using attribute reduce theory of RS to obtain no redundant attributes, and detect shot by using the theory of classification of RS. Finally, the performance of the algorithm is K-mean cross validation. Experimental results show that the algorithm can reduce the computational complexity and improve the accuracy of the shot detection effectively.