Video Copy Detection based on Local Ordinal

The ordinal is a popular method used in video copy detection. In this work,an efficient content-based copy detection approach is proposed with novel ordinal features. The features are obtained from intensity relation between adjacent blocks of each frame which obey Hilbert curve order and used to generate hash bits. To effectively localize a short query video clip in a long target video,hash comparison scheme is developed. In the scheme,sequence similarity rate is proposed to handle similarity matching,and dynamic programming is applied to improve accuracy. Detection samples are constructed for comparing the performance of the proposed approach with the traditional ordinal methods. It is showed that the proposed approach has better performance in accuracy and suitable for video copy detection.