An Approach for Video Summarization Using Graph-Based Clustering Algorithm

There has been immense increase in amount of video content over the past few years. This massive growth in the content of video leads to the uncertain outcomes. Processing of these huge chunks of data is demanding plenty of resources, like time, manpower, as well as hardware storage. Video summarization acquires an important remark in this ambiance. It supports in providing efficient storage, fast browsing, and retrieval of huge video data without losing important factors. In this work, video summarization technique has been proposed using graph-based clustering algorithm in three different steps. The presented work has split the video into frames of a predefined time period and computed the similarity between consecutive frames. Based on the similarity values, the frames are grouped into scenes, and the scenes are partitioned separately using video tracks and audio tracks with the help of clustering algorithm. Then the combined clusters of scenes are further analyzed to determine the summary of a video file. This work considers just the video file at hand and attempts to develop a summary file without using any external knowledge from similar videos. As per the authors’ knowledge, no previous research work related to video summarization has been conducted in this field. The quality of the summarized video is a measure to express the effectiveness of the proposed methodology.