SIFT and DWT based Video Sequence Matching Method for Videocopy Detection

Abstract

This paper introduces SIFT and DWT based video sequence matching method for video copy detection. Since local features have good stability and discriminative ability SIFT descriptors are used here for video content description. Content matching using SIFT takes large amount of time and it is computationally expensive for high dimensions and large number of points. These difficulties are solved by using dual threshold method which divides videos into segments having homogeneous content and by performing keyframe extraction on each of these segments. SIFT features are then extracted from these keyframes and SIFT feature sets of two video frames are matched using SVD based method. It has the problem of high processing time proportional to the length of video content. So we proposed DWT based fingerprint generation technique to reduce the processing time. Fingerprints of videos are generated and fingerprint matching is performed in the preprocessing step. So based on these results, it decides whether the SIFT feature matching has to be performed or not. Experimental results shows that SIFT and DWT based video sequence matching method for video copy detection can effectively detect video copies. Proposed system has following advantages such as, based on the spatial features it can effectively find optimal sequence matching result from the disordered matching results, it can effectively reduce the processing time and it is adaptive to video frame rate changes. Experimental results also demonstrate that the proposed method can obtain a better tradeoff between the effectiveness and the efficiency of video copy detection. KeywordsVideo copy Detection, Scale Invariant Feature Transform, Singular Value Decomposition, Discrete wavelet transform.