Automatic video parsing using shot boundary detection and camera operation analysis

摘要

In this paper, we present an efficient video parsing method for automating content-based video indexing and retrieval using shot boundary detection and camera operation analysis techniques. In the shot boundary detection, the local color information is used in order to eliminate the false detection caused by an abrupt change of illumination such as camera flash or thunder. In order to reduce the computation time in the shot boundary detection, an adaptive time window is applied to this procedure. Local spatio-temporal images and multilayer perceptron are used for analyzing camera operations. The proposed method uses a learning algorithm with spatio-temporal information in the frame and does not process the entire video image to reduce the processing time. In order to verify the performance of the proposed automatic video parsing method, experiments have been carried out with a video database that includes news, documentary and movie. Experimental results demonstrate the efficiency of the proposed video parsing technique.