Key Frame Extraction Based on Shot Coverage and Distortion

Abstract

Key frame extraction has been recognized as one of the important research issues in video information retrieval. Until now, in spite of a lot of research efforts on the key frame extraction for video sequences, existing approaches cannot quantitatively evaluate the importance of extracted frames in representing the video contents. In this paper, we propose a new algorithm for key frame extraction using shot coverage and distortion. The algorithm finds significant key frames from candidate key frames. When selecting the candidate frames, the coverage rate for each frame to the whole frames in a shot is computed by using the difference between adjacent frames. The frames with the coverage rate within 10% from the top are regarded as the candidates. Then, by computing the distortion rate of a candidate against all frames, the most representative frame is selected as a key frame in the shot. The performance of the proposed algorithm has been verified by a statistical test. Experimental results show that the proposed algorithm improves the performance by 13 – 50% over the existing methods.KeywordsVideo SequenceOptical FlowCoverage RateMultimedia ContentDistortion RateThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.