Change detection with automatic reference frame update and key frame detector

In this paper a new reference image approach for real world surveillance applications is presented. The presented algorithm is composed of two steps. In the first step, A statistical test is performed between a reference and a current image, and a decision is made whether there is a change at a given pixel position. In the second decision step, the possibility to update a reference frame is provided. The second step allows to use the the proposed overall algorithm as a surveillance key frame extractor. These key frames are indicating significant changes with respect to a previously stored reference frame as well as to dynamically updated reference frames chosen from the ongoing sequence.