Abstract: Image distortion is a difficult problem in single grayscale specular images detection and removal method, so based on the mean field annealing algorithm a specular detection method of grayscale image is presented. The specular and diffuse reflection components are modeled by mean field annealing algorithm. Then, proportion of the specular and diffuse reflection components is estimated. Possible specular area is detected with proportion. Finally, the image restoration method based on BSCB model removals highlights. Experimental results show that the proposed method is effective. Compared with traditional highlight detection and recovery methods, this method can effectively detect grayscale image specular reflection area. The more the effect is restored in line with human vision, the better the image quality is recovered, thus the method improves the recovery rate of the image highlight areas.