Using perceptual attributes to obtain dense depth maps

An efficient technique is presented for obtaining dense stereo correspondence using a new adaptive block matching technique. First, one of the color stereo images (e.g., the left) is segmented into chromatic and achromatic regions using perceptual attributes in the HSI color space. Then the block matching technique is applied to the stereo images using adaptive similarity measures. The results have been compared to the results of the gray value algorithm. The precision of the matching results always improved by 25 to 30% when using the adaptive color algorithm instead of the gray value algorithm.