Moving cast shadow suppression from a Gaussian mixture shadow model

Abstract

The background modeling of complex environment, moving object detection, and moving cast shadow suppression could be applied to a lot of fields such as intelligent surveillance, robot vision and videoconference ect. In moving foreground detection, an improved mixture Gaussian-based background modeling method was presented, which updated the parameters of Gaussians according to the frequency of pixel value changes, to reduce the cost of computation. In shadow suppression, a mixture Gaussian-based clustering algorithm was provided to detect and suppress shadow. This method firstly identifies whether a pixel value is probable shadow by shadow model in HSV color space, the pixel values detected as probable shadow are then put into mixture Gaussian shadow model to learn and cluster. The experimental results indicates that the proposed approach in this paper can process in real-time and remove shadow more effectively.