Foreground detection based on color and texture features

Moving objects detection from a video sequence is a fundamental and critical work in many computer-vision applications. A widely utilized strategy is to implement background subtraction method, aiming to classify moving objects from video sequences based on the background model. To address the false classification, a new background modeling method is presented and a new foreground object detection model is proposed which utilize the proposed background similarity to fuse color and texture feature. It can tackle relatively complex scenarios with small motions and the illumination changes. The experimental results have proven the effectiveness of the method.