A feature-based stereo model using disparity histograms of multi-resolution channels
Yoshiro Nishimoto, Yoshiaki Shirai · Advanced Robotics · 1988
A feature-based stereo model designed to perform matching multi-resolution features is described. The multi-resolution features are zero-crossings (ZCs) of images convolved with different sized Laplacian-Gaussian operators. ZCs corresponding to small contrasts are removed according to the gradient values of images convolved with Gaussian operators. Candidate disparity intervals are determined using a disparity histogram of the total ZCs over the entire image. The image is then divided into small areas and a local disparity histogram is computed for the candidate intervals. Local disparity histograms in all the resolution channels are searched for the most promising disparity in each area. If the disparity is found successfully, the disparities for all the ZCs in the area are determined by searching only for the neighbour of the promising disparity. Once a disparity for a ZC is determined, the matching pair of ZCs is removed from the set of ZCs. This process is iterated until no more disparities are determined. Experiments with sample scenes including objects of various shapes and brightness at different positions reveal that the model has advantages in efficiency and performance. A discussion is also given on the correspondence between the stereo model and the human binocular system.