Image Segmentation and Geometric Feature Based Approach for Fast Video Summarization of Surveillance Videos

In this paper, we propose a geometric feature and frame segmentation based approach for video summarization. Video summarization aims to generate a summarized video with all the salient activities of the input video. We propose to retain the salient frames towards generation of video summary. We detect saliency in foreground and background of the image separately. We propose to model the image as MRF (Markov Random Field) and use MAP (Maximum a-posteriori) as final solution to segment the image into foreground and background. The salient frame is defined by the variation in feature descriptors using the geometric features. We propose to combine the probabilities of foreground and background segments being salient using DSCR (Dempster Shafer Combination Rule). We consider the summarized video as a combination of salient frames for a user defined time. We demonstrate the results using several videos in BL-7F dataset and compare the same with state of art techniques using retention ratio and condensation ratio as quality parameters.