Comparison of distance measures for video copy detection
Arun Hampapur and Ruud Bolle
submitted to International Conference on Multimedia and Expo 2001 (ICME-2001), 2001.
(also IBM Research Report, Computer Science, RC 22056, May 2001.)
This paper compares the performance of a number of image distance measures (color histogram intersection,image difference, edge matching, edge orientation histogram intersection, invariant moments and Hausdorff distance) for comparing video frames for the purpose of video copy detection. The original images are extracted from MPEG1 video with 352 x 240 resolution. The copies are derived from AVI video with 240 x 180 resolution. The video is taken from cable television commercials. The image copies are matched against each other using each of the distance metrics and the performance is measured using the false positive and false negative rates.