Making Background Subtraction Robust to Various Illumination Changes

Background subtraction is an important step used to segment moving regions in surveillance videos. Modern background subtraction techniques can handle gradual illumination changes but can easily be confused by rapid ones. In particular, varying illuminations cause significant changes in the representation of a scene in different spaces, which in turn results in the high levels of failure in such conditions. Especially, sudden illumination changes often falsely labeled as foreground objects, which may severely degrade the accuracy of object localization and detection. Thus in this paper, we propose a robust background modeling technique that overcomes this limitation by employing adaptive-length recursive hybrid median filters. This algorithm can achieve significantly better image quality than fixed length standard median filters when the images are corrupted by impulsive noise making our approach extremely robust to illumination changes, whether slow or fast. The performance of the proposed algorithm is compared with slow median filters, rapid median filters, and hybrid median filters by showing its effectiveness for occlusion handling in real time scenarios.