Empirical Comparison of Different Key Frame Extraction Approaches with Differential Evolution Based Algorithms

Key frame extraction is an integral part of video analytics. The extracted key frames are used for video summarization and information retrieval. There exist many approaches for solving key frame extraction problem in video analytics. The focus of this paper is to extend the strategy of integrating Evolutionary Computing technique with a conventional key frame extraction approach, which is proposed by the authors in their previous work, with two other conventional approaches. The conventional approaches considered in this study are SSIM (Structural Similarity Index Method) Method, Entropy Method and Euclidean Distance method. This paper also proposes a new approach for key frame extraction by integrating the Euclidean Distance method with Differential Evolution algorithm. The proposed approach is compared with all the existing approaches by its speed and accuracy. It is found from the comparison that the proposed approach outperforms other approaches. The results and discussion related to this experiment study are presented in this paper.