Robust background modeling based on multiscale color description

Background subtraction is the basis of object detection and tracking for machine vision systems. Traditional background modeling methods often require complicated computations and are sensitive to illumination changes and shadow interference. In this paper, we propose a multiscale background modeling method, which fully utilizes the color characteristics of each incoming frame. The proposed method is quite efficient and is capable of resisting illumination changes and shadow disturbance. Experimental results show that our method is suitable for real-world scenarios and real-time applications.