Straightforward Random Noise Model for Image Registration in Binocular Systems

摘要

This paper proposed a straightforward random noise statistical model based on the stereo matching characteristic in binocular systems. The noises of affecting stereo matching precision are classified into three categories: measuring noise, illumination noise and noise coming from parallax. Our theoretical analysis and experiment show that the noises of affecting stereo matching precision have a local Gaussian distribution. Based on the noise model, an algorithm to reduce noise while preserving useful information is proposed. Experimental results show that the algorithm has the advantage of robust and can make 3D measuring precision of binocular systems better than +-1/4 pixel in practical applications.