ABSTRACTLocal depth variation is a distinctive property of natural scenes and its effects on perception have only recently begun to be investigated. Here, we demonstrate how natural depth variation impacts performance in two fundamental tasks related to stereopsis: half-occlusion detection and disparity detection. We report the results of a computational study that uses a large database of calibrated natural stereo-images with precisely co-registered laser-based distance measurements. First, we develop a procedure for precisely sampling stereo-image patches from the stereo-images, based on the distance measurements. The local depth variation in each stereo-image patch is quantified by disparity contrast. Next, we show that increased disparity contrast degrades performance in half-occlusion detection and disparity detection tasks, and changes the size and shape of the optimal spatial integration areas (“receptive fields”) for computing the task-relevant decision variables. Then, we show that a simple binocular image statistic predicts disparity contrast in natural scenes. Finally, we report results on the most likely patterns of disparity variation in natural scenes. Our findings motivate computational and psychophysical investigations of the mechanisms that underlie disparity estimation in local regions of natural scenes.
Geometrically constrained sub-pixel disparity estimation from stereo images of the retinal fundus
