Storyboard Based Video Browsing: Optimizing human visual inspection for video archive navigation and search

Abstract

Abstract With the size increase of video databases it is important to create an efficient video browsing system. We investigate how to optimize the interface, relying solely on human visual inspection. This thesis sets its focus on the latter because research has shown the tremendous power of such human abilities. Yet, despite the resulting potential advantages, these facts are often neglected in current system designs. We go into the fundamentals of a storyboard system which has been proven to perform well in the VBS 2015 by testing the size and layout of its frames, and research the contribution of the storyboard system, when integrating it with a video browsing system relying on filtering of results. The results yielded statistically significant differences between certain size and layout combinations, while the contribution test yielded less strong, yet positive evidence. Suggestions for future work consist of researching variables that could improve the storyboard interface, such as the color contrast, that could enhance object recognition as well as ways to optimally integrate filtering with an intelligent interface by testing the effects of different filters as well as experimenting with different frame rates.