An Effective Background Subtraction under a Continuosly and Rapidly Varying Illumination

Abstract

Background subtraction is an effective method in detecting moving objects in a static scene, which requires a fixed camera, and a static background. Illumination changing is a challenging problem which causes failure of background subtraction. Most background subtract algorithms require the illumination changing slowly, so the object can be tracked accurately and the background is easy to update. If the illumination changes rapidly and nonmonotonously, many algorithms fail. In this paper, we do some research on the condition when there is a single light which simulates a continuously and rapidly varying illumination. This kind of condition has many applicable scenes in daily life. This scene can be seen not only in doors or in the dark or in underground architecture but also out doors or in the light. We propose a new algorithm based on the illumination factor which can be applied to not only the indoor environment but also outdoors with natural illumination. In our algorithm, illumination factor is computed through the comparison of background and the current frame. The background is estimated and updated based on this factor, which tracks the change of background accurately and reduces the chance of failure. Experiments have proven the effectiveness of our method.