Near-Duplicate Web Video Retrieval and Localization Using Improved Edit Distance

With the development of network, there exists many near-duplicate videos online shared by individuals. These ones cause problems such as copyright infringement and search result redundancy. To solve the issues, this paper proposes a filter-and-refine framework for near-duplicate video retrieval and localization. By regarding video sequences as strings, Edit distance is used and improved in the approach. Firstly, bag-of-words (BOW) model is utilized to measure the similarities between frames. Then, non-near-duplicate videos are filtered out by computing the proposed relative Edit distance similarity (REDS). Next, a dynamic programming strategy is proposed to rank the remained videos and localize the similar segments. Experiments demonstrate the effectiveness and robustness of the method in retrieval and localization.