Background differencing technique for image segmentation based on the status of reference pixels

Background frame differencing is one of the most efficient, image segmentation techniques in case that a static background frame is available. The technique, however, involves two major problems: one is that decision with a single global threshold often produces poor segmentation results and the other is that an estimation for the optimal threshold is actually difficult. To overcome these problems, this paper proposes a very simple yet effective algorithm that refers to the status of previously judged neighboring pixels: more specifically, the proposed algorithm varies the threshold value in accordance with the status of the reference pixels. The experimental results illustrate that the proposed technique gives better segmentation results than the simple background differencing one with the optimal global threshold.