A Video Similarity Matching Algorithm Supporting for Different Time Scales

Based on analysis on video retrieval issues among different time scales, a dynamic time warping (DTW) optimal matching algorithm is proposed in this paper. Firstly the sub-segments detection algorithm was used to separate the video into several consecutive sub-segments; furthermore, the video similarity matching algorithm on the basis of sub-segments was used to implement the retrieval process. The proposed algorithm establishes sub-segment correspondences by minimizing global differences between them, introduces search bound to change the whole searching to local searching, keeps temporal order of sub-segments and handles nonlinear misalignment of videos. It not only can successfully matching similar videos which have different time scales, but also meet the demand of rapid retrieval among massive video database. Experimental results show that the proposed algorithm is more efficient and robust than the conventional similarity matching algorithm based on visual similarity.