Network-aware identification of video clip fragments
José San Pedro, Sergio Domínguez · 2007
In this paper, we present a visual hash-based video identification system able to work under variable video encoding conditions, a common scenario in network-aware applications, and specifically targeted to the detection of partial appearances of videos. The algorithm extracts video entropy, and based on the resulting series, a set of intervals and frames for each interval are selected. The same exact intervals and frames can be repeatedly extracted for any other variation of the original content, which allows to perform identification when variations have undergone dramatic quality and frame rate reduction in their distribution. The problem becomes much more challenging when dealing with fragments of the complete content; by loosing an a-priori reference point (e.g. first frame of the video) string comparison techniques fail to choose the same subset of frames to generate the identification fingerprint, which leads to a significant decrease in recall values. Differently from basic subsampling techniques, the proposed method is able to get a non-referenced frame set appropriate for its use under the mentioned conditions.