Video Story Board Design

Video Story Board Design

Abstract :

Rapid advances in data storage, data compression and data transmission has led to the emergence of an important research area in the recent times known as Video Summarization. Video storyboard, a set of static key frames preserving the overall content of the video with minimum data, enables an user to access any video in a friendly and meaningful way.

In this paper, a novel automated method for construction of video storyboards based on information-theoretic pre-sampling and Delaunay graph-based clustering is proposed. More informative frames are included during the pre-sampling stage based on detection of significant valleys in the mutual information profile of successive frames of the video. This strategy reduces the chance of loss of information to a great extent as compared to only fixed pre-sampling. Robust clustering is achieved with a judicious edge pruning strategy via maximum global standard deviation reduction of edge lengths with a constraint on deviation ratio of the vertices in the Delaunay graph. A comprehensive evaluation is being done on 100 videos, 50 each from Open Video Project and You Tube, using both objective and subjective measures clearly demonstrates our method outperforming three state-of-the-art video summarization techniques.

A sample sheet of user survey and our results for all the 100 videos are provided below.