Multilevel Framework for Summarization of Surveillance Videos

In this paper, we propose multilevel framework for summarization of surveillance videos using motion entropy by maintaining chronology of activities. We aim to reduce the size of the video to give a meaningful summary by retaining important activities without destroying the temporal relationship of the activities. Initially input video is divided into blocks and then blocks into segments in a multilevel framework with a non-uniform number of frames in a block at each level. We present a mechanism to select salient blocks which can contribute in the video summarization by propagating motion information from frame level to segment level and further to block level. We generate the summarized video at each level of the framework and is stopped at a level where optimization of condensed ratio between successive levels is minimal. The experimental results on surveillance videos shows better condensed ratios.