Entropy Based Fuzzy C Means Clustering and Key Frame Extraction for Sports Video Summarization

Abstract

Recent advances in technology have made tremendous amount of multimedia information available to the general population. To access the needed information in this scenario there is a need for automatic tools to filter and present information summary. Summarization techniques will give a choice to users to browse and select the multimedia documents of their choice for complete viewing later. In this work a new summarization technique to collect frames of importance in a video is presented. The method is based on selection of frames typically different from their immediate neighbors as key frames from group of similar frames. It uses the process of clustering, where visually similar frames are collected into one group using Fuzzy C means clustering algorithm. When clusters are formed, the frames that exhibit a change ratio which is a measure of the content variation, greater than the average value of the cluster are treated as Key frames. The summary is created by merging Key frames on the basis of their timeline. This method ensures that video summary represents the most unique frames of the input video and gives equal attention to preserving continuity of the summarized video. The robustness of the algorithm is validated by average values of performance parameters. The average compression ratio of 92% is indication of higher conciseness. The average fidelity of 95% is an indicative of comprehensive representation of video by the key frames selected using proposed algorithm.