In order to achieve real time human posture extraction from image, this paper proposes a method of background subtraction using color information of each pixel. In the proposed method, the threshold processing of both brightness difference and color difference between an averaged background image and a current image is carried out. A weight function is also introduced to define the threshold when the pixel has low level resolution.
Experimental results show that the proposed method can extract a target object from image without controlling the background and is more effective than the conventional background subtraction method.