Video retrieval method using non-parametric based motion classification

In this paper, we propose a novel video retrieval method using nonparametric based motion classification in the shot-based video indexing structure. The proposed system gets the representative frame and motion information from each shot segmented by the shot change detection method, and extracts visual features and non-parametric based motion information from them. Then, we construct a real-time video retrieval system using similarity comparison between these spatio-temporal features. The extraction of non-parametric based motion features, after the normalized motion vectors are created from an MPEG-compressed stream, is effectively fulfilled by discretizing each normalized motion vector into various angle bins, and by considering the mean, variance, and direction of motion vectors in these bins. In addition, we use the edge-based spatial descriptor to extract the visual feature in representative frames. Experimental evidence shows that our algorithm outperforms other video retrieval methods for image indexing and retrieval. To index the feature vectors, we use R*-tree structures.