An Improved Video Identification Scheme Based on Video Tomography.
Qing-Ge Ji,Zhi-Feng Tan,Zhe-Ming Lu,Yong Zhang
DOI: https://doi.org/10.1587/transinf.e97.d.919
2014-01-01
IEICE Transactions on Information and Systems
Abstract:In recent years, with the popularization of video collection devices and the development of the Internet, it is easy to copy original digital videos and distribute illegal copies quickly through the Internet. It becomes a critical task to uphold copyright laws, and this problem will require a technical solution. Therefore, as a challenging problem, copy detection or video identification becomes increasingly important. The problem addressed here is to identify a given video clip in a given set of video sequences. In this paper, an extension to the video identification approach based on video tomography is presented. First, the feature extraction process is modified to enhance the reliability of the shot signature with its size unchanged. Then, a new similarity measurement between two shot signatures is proposed to address the problem generated by the original approach when facing the query shot with a short length. In addition, the query scope is extended from one shot only to one clip (several consecutive shots) by giving a new definition of similarity between two clips and describing a search algorithm which can save much of the computation cost. Experimental results show that the proposed approach is more suitable for identifying shots with short lengths than the original approach. The clip query approach performs well in the experiment and it also shows strong robustness to data loss.