A Study on Multi-view Video Summarization Techniques

Abstract

With the recent developments in multimedia and their applications in day-to-day life, the summarization of video content is a growing and popular field. The main objective of video summarization is to abstract the important events, or objects from a video to provide an easily interpretable summary. The video-surveillance is a crucial security mechanism of many public and private places like hospitals, universities, banks, etc. These videos provide important information about any intruder or invaders. Although this information is of critical importance, it takes a lot of time and effort to go through hours of long videos in order to find out the significant events. Here, multi-view video summarization comes to rescue as it extracts the most useful events out of the videos and presents it to the user. The multi-view video summarization system provides an advantage over a single view video summarization system by exploiting the complicated intra-view and inter-view correlations, thereby presenting an effective summary of multiple camera views and enriching the video content. It explores various dimensions of the same event and collectively may give more complete information. In this paper, we present a study on multi-view video summarization techniques. We examine work done on various methods of video summarization for videos consisting of many views of the same location as captured by the surveillance cameras positioned at different angles. The final summarized video can be used for the identification of suspicious events, an inspection of post-accidental events, thefts in public places, etc., which will be supportive to the crime department for investigation purposes.