A Two-Stage Intelligent Compression System for Surveillance Videos

Surveillance videos are becoming immensely popular nowadays due to the increasing usage of surveillance systems in various places around the world. In such applications, cameras capture and record information over long durations of time, which result in large quantities of data, necessitating specialized compression techniques. Conventional video compression algorithms are not sufficient and efficient enough for such videos. In this paper, a novel two-stage compression system for surveillance videos is proposed that can automatically adapt the compression based on the semantic content of the video data. The initial stage consists of an “intelligent interesting event detector” that discards groups of frames in which no interesting events are detected, effectively reducing the size of the video without any loss in video quality. The removal process is robust to minor illumination variations and other small periodic movements. In the second stage, the remaining frames are compressed by HuffYUV codec which is a lossless compression scheme. Results indicate that compression ratios that can be achieved by our system are very encouraging and we demostrate the effectiveness of our system on seven different surveillance videos consisting of a wide range of scenerios. Key–Words: Intelligent system, Video processing, Compression, Surveillance system