On Stereo Correspondence Estimation: A Spiral Search Algorithm

This paper introduces a searching algorithm called "spiral searching" for computing stereo correspondence or disparity of the stereo images. The method is based on computation of the minimum window cost among the contributions of the windows bounded in the range from minimum depth of spiral to maximum depth of spiral. This algorithm can estimate stereo correspondence of a pair of images concurrently two dimensionally and it avoids false matching causes to increase the accuracy and requires minimum executing time than the traditional one dimensional searching strategies. This method first calculates two window costs - one in positive x-direction and another in negative y-direction using the same distance from the origin. Minimum of the two window costs and coordinate distances are considered for second calculation. Secondly, following the same way another two window costs are calculated - one in negative x-direction and another in positive y-direction using the same distance from the origin. Minimum of the two window costs and coordinate distances are compared to the previous two window costs. This process is bounded from minimum depth of spiral to maximum depth of spiral. Experimental result demonstrates that the visual quality of the output image is very close to ground truth image.