Fast stereo correspondence using small-color census transform

This paper presents small color census transform for disparity estimation which is developed based on perceptual color contrast. Color distance between two color points corresponds to the perceptual difference between two colors in human vision. The color distance can be used as input in the census transform. The small census transform idea is adopted from mini census transform in grayscale. Different color spaces are implemented to see the performance. The experimental result shows that original census transform, mini census transform and small color census transform using sum hamming distance (SHD) can significantly reduce the processing time compared to sum of absolute differences (SAD). Small census transform takes more time compared to grayscale mini census transform, however it is not significant. Even though small color census transform takes more time compared to mini census transform, it is considerable to be used for initial disparity map computation.