An image segmentation technique which incorporates spatial (contextual) information is described in this paper. This method uses the sigma probability concept to determine if the center pixel of a window is to bc merged with a previously established region within the window or if it is to become the initial member of a ncw rcgion. Also. most of the pixels in this sigma range must be connected: this not only ensures that those randomly distributed noise pixles can be avoided but also ensures the full utilization of contextual information from the image. Experimcntal rcsults are provided to show the effectiveness of the proposed segmentation method.