On the basis of imaging characteristics of a reflection stereo vision, an approach for disparity estimation and image restoration of reflection stereo vision was proposed and the double image of the vision system was restored.The reflection stereo vision is a monocular system, and can capture the double image in real time. In experiment, the image was segmented into the rectangle observing blocks firstly, then, the cepstrum analysis was used to estimate the displacement amount (disparity) of each pixel in the double image by detecting the two symmetrical peaks from the cepstrum domain and to calculate the 3D space distance of a target point. Finally, the disparity was used to combine an inverse filter to restore the double image and to obtain a good scene image. Experimental results show that the method is effective for disparity estimation and image restoration, and the accuracy of the disparity estimation has improved by 15%-20%.