VISUC: Video summarization with user customization

Abstract

Design of video storyboards, which enables a user to access any video in a friendly and meaningful way, has emerged as an im p ortant area of research in the multimedia community. In this p a p er, we p ro p ose a novel semi-automated method for construction of video storyboards based on Delaunay gra p hs. A robust edge p runing strategy, where the edge weights are assumed to follow a Gaussian distribution, is a pp lied on an a pp ro p riately constructed Delaunay gra p h. The p ro p osed method also takes into account two advanced user needs, namely the waiting time and the number of frames an user wants to see in the storyboard. Ex p erimental results on some standard videos of diff erent genre clearly indicate the su p eriority of the p ro p osed method in terms of the FO.5 measure.

Video Key Frame Extraction through Canonical Correlation Analysis and Graph Modularity

“…VISON). In order to verify the statistical significance of those results, the confidence intervals for the differences between paired means are computed [5][6]10]. Since the confidence intervals (with a confidence of 98%) do not include zero in both the cases (from Table 2), the results presented in Table 1 confirm that proposed method outperforms both VISON and VSUMM in statistically significant manner.…”

“…In this work, we evaluated our technique based on the Comparison of User Summaries (CUS) mechanism proposed by Avila et al [5]. This evaluation scheme is also adopted by various video summarization approaches for performance comparison [3,[5][6]10]. In this evaluation scheme, the manually created user summaries are taken as references for comparison with the summaries generated by the automated methods.…”

“…In this evaluation scheme, the manually created user summaries are taken as references for comparison with the summaries generated by the automated methods. The performance was validated by F 1 -measure which is a function of both precision and recall [6,10].…”

“…Different clustering techniques have been proposed in the literature to address the problem of extracting key frames from a video sequence [4][5]10]. The performance of these methods heavily depends on user inputs and/or certain threshold parameters [5].…”