Practical on-line near-duplicate detection for web video

This paper proposes an alternative point of view on the use of ordinal measures to detect near-duplicate web videos. Ordinal measure fingerprints are first extracted in the core area of videos, and ordinal measures are fast matched based on Transformation distance, and then an asymmetric technique is used to obtain the longest same sub-sequence and detect video copies. Experimental results show that the method has good robustness and distinctiveness.