Foreground objects detection using multiple difference images

摘要

Abstract. In visual surveillance, robust foreground object detection is annessential step for further processing such as segmentation, tracking, andnextraction of a scene’s contextual information. Typical approaches con-ntinuously update background images and use then for detecting fore-nground objects. They involve many parameters that should be adjustednaccording to the situation where surveillance cameras are operating. Wenpropose an algorithm for the robust detection of foreground objects usingnmultiple difference images that requires only one parameter to adjust.nWe show that the proposed algorithm gives comparable results with lessncomputation time through experimental results using test images withngroundtruths. © 2010 Society of Photo-Optical InstrumentationnEngineers. u0001DOI: 10.1117/1.3374043