Video shot boundary detection using statistical methods

Automatic video shot boundary detection is one of the major challenges involved in video piracy detection. In this paper, we propose a novel approach for detecting boundaries of the shots embedded within a video stream. Initially, video stream is segmented into different frames. Each frame will undergo preprocessing technique which will ease the further process of feature extraction. Statistical features like mean and standard deviation are computed for the frame differences. Local maxima of the vector of averages are considered to identify the outlier frame where there is a possible boundary of one shot. This method is found to be computationally simple and efficient. The accuracy of the proposed algorithm is found to be 98.9939%.