Robust Hashing Based on Persistent Points for Video Copy Detection

Abstract

Web video databases tend to contain immense copies with the explosive growth of online videos. How to detect video copies has become an essential problem for web video storage and indexing. This paper presents an efficient Hash scheme for video copy detection. By building trajectories of Harris points from frame to frame, behavior of interest points can be expressed. Spatio-temporal persistent points are selected to get differential description of local region around as local description features. An indexing method is used to generate final hash from the features, which can index multi-dimensional data to one dimension. To effectively localize a short query video clip in a long target video through the proposed video hash, we developed a hash comparison scheme. Experiments showed that the proposed hash scheme has excellent performance in accuracy and speed.