Video Summarization Based on Optical Flow

Abstract

The explosive growth in digital videos demands a technique that effectively identifies informative parts from the video. Video summarization refers to creating a video summary as a collection of keyframes that depicts key actions and events in the video. The authors propose to generate a video summary based on apparent motion information in the video, that is, optical flow. The proposed algorithm uses optical flow technique to estimate the change in the local flow of pixel intensities to identify the keyframes. The proposed algorithm is tested on two standard databases, such as Open Video Project and YouTube database. The results and the quantitative evaluation validate the effectiveness of the proposed algorithm for generation of a video summary.