A robust local census-based stereo matching insensitive to illumination changes

In this paper, we present a novel robust census transform stereo matching algorithm based on the classic census transform. While many stereo algorithms have been proposed in recent years, their similarity measures greatly depends on the intensity statistic of the stereo image pairs and are sensitive to the noise. Although the classic census non-parametric transform can improve the performance of disparity maps in the nonideal illumination condition, it also has some drawbacks. We take the summation of the intensity mean and a small value as the center pixel to revise the classic census transform. Experiments have been carried out under different illumination conditions, and the results show that the proposed algorithm achieves high efficiency and accuracy and is robust to the illumination changes.