An evolutionary computing approach for solving key frame extraction problem in video analytics

Key frame extraction is a very important task in video analytics. The extracted key frames are used for video summarization and event detection for the given videos. This paper proposes a new algorithmic framework which combines the conventional SSIM (Structural Similarity Index) approach with the state-of-the-art Differential Evolution (DE) algorithm for key frame extraction from video data. The proposed algorithm is implemented to extract key frames from three different videos and its performance is compared with the conventional approach using the average SSIM values. The results show that DE based approach gives comparatively better results when compared to the naive SSIM approach.