No Detail Found!

A new detection algorithm for video cut using a local threshold.

Author(s)
Hun-Woo Yoo Image Information System Lab, Department of Industrial Systems and Information Engineering, Korea University, Sungbuk-gu Anam-dong 5 Ga 1, Seoul 136-701, Korea
Dong-Sik Jang Image Information System Lab, Department of Industrial Systems and Information Engineering, Korea University, Sungbuk-gu Anam-dong 5 Ga 1, Seoul 136-701, Korea
Abstract
A new statistical approach for the detection of shot boundaries is proposed in this paper. This method operates directly on MPEG compressed videos. An intersection of the DC luminance histogram is used as a similarity metric of inter-frame. A confidence interval computed from inter-frame similarity values in a local MPEG sequence window is used as a local threshold to detect abrupt shot boundary (cut). This method adaptively varies a threshold according to the statistical properties of video sequences. These time varying characteristics can eliminate detection errors, which frequently occur in a global threshold approach. Experimental results show that the proposed algorithm has more than an average 96 % recall and 72% precision for selected video sequences.
Publication Details
Page(s) 267-273
DOI DOI not available
Published Journal: Asian Journal of Information Technology, Volume: 2, Issue: 4, Year: 2003

Request the full text from PASTIC

More from Asian Journal of Information Technology
0

Citations

0

Downloads

122

Views

Copyright ©  PASTIC National Center, Islamabad www.pastic.gov.pk