A Segmentation Based Sequential Pattern Matching for Efficient Video Copy Detection

We propose a video copy detection system which is sufficiently robust to detect transformed versions of original videos with ability to pinpoint location of copied segments. Precisely due to the good stability and discriminative capability, SIFT feature is used for visual content description. However SIFT based feature matching has high computational cost due to large number of keypoints and high dimensionality of its feature vector. Thus to minimize computational complexity, video content is divided into number of segments depending on homogeneity of video content. SIFT features are extracted from keyframes of video segments. Then SIFT features are quantized to generate clusters. Further binary SIFT are generated for every cluster for optimized matching. To perform video segment matching with SIFT descriptors, firstly visual words matching is done at cluster level then at feature level, similarity measure is employed. In order to find video copy detection, an optimal sequential pattern matching is done which will ensure effectiveness and efficiency of the proposed video copy detection system. Our video copy detection system gives improved results in terms of accuracy and has been able to detect most of the dataset query video transformations in time-efficient manner. Keywords— Copyright protection, content based video copy detection, feature extraction, visual features, SIFT feature.