Robustness improvement for the calibration of stereo deflectometry based on a search algorithm

Iterative calculation is a necessary step in the calibration of stereo deflectometry. Inaccurate input can result in the iterative process converging in a wrong position or unconvergence. Image distortion is an important factors affecting the accuracy of the input. In order to reduce the influence of image distortion and increase the robustness of the calibration, a method based on a search algorithm is investigated for stereo deflectometry. Because there is few distortion at an image center, a search window with a certain border length is positioned at the image center to obtain a group of data for the iterative process. The size of the window is determined based on an algorithm proposed in this paper. Due to the fact that the centers of distortion and image are not coincident, the window is consecutively relocated within the image. A function is proposed to evaluate the input accuracy. Along with the window moving, the calculated data which makes the proposed function reach the minimum is selected to compute the following iterative process. Experimental results affirm the presented method can significantly enhance the robustness of the calibration accuracy of a stereo deflectometry system. By applying the proposed method, the RMS (root mean square) of calibration error can be increased from 0.31 pixels to 0.05 pixels.