Stereo Correspondence Estimation Based on Wavelets and Multiwavelets Analysis

Abstract

The developed vision system consists of a novel robust algorithm. The proposed algorithm uses the stereo vision capabilities and multi-resolution analysis to estimate disparity maps and the concerned 3D depths. Furthermore, it uses multiwavelets theory that is a newer way of scale space representation of the signals and considered as fundamental as Fourier and a better alternative. The proposed algorithm uses the well-known technique of coarse-to-fine matching to address the problem of stereo correspondence. The translation invariant wavelets transform modulus maxima (WTMM) are used as corresponding features. To keep the whole correspondence estimation process consistent and resistant to errors, optimized selection criterion strength of the candidate is developed. The strength of the candidate involves the contribution of probabilistic weighted normalized correlation, symbolic tagging and geometric refinement. Probabilistic weighting involves the contribution of more than one search spaces, whereas symbolic tagging helps to keep the track of the most significant and consistent candidates throughout the process. Furthermore, geometric refinement addresses the problem of geometric distortion between the perspective views. The geometric features used in the geometric refinement procedure are carefully chosen to be invariant through many geometric transformations, such as affine, metric, Euclidean and projective. Moreover, beside that comprehensive selection criterion the whole correspondence estimation process is constrained to uniqueness, continuity and smoothness. A novel and robust stereo vision system is developed that is capable of estimating 3D depths of objects to high accuracy. The maximum error deviation of the estimated depth along the surfaces is less than 0.5mm and along the discontinuities is less than 1.5mm. Similarly the time taken by the algorithm is with in the range of [12-15] seconds for the images of size [640-480]. The proposed system is very simple and consists of only a stereo cameras pair and a simple fluorescent light. The developed system is invariant to illuminative variations, and orientation, location and scaling of the objects, which makes the system highly robust. Due to its hardware simplicity and robustness, it can be implemented in different factory environments with out a significant change in the setup of the system. Due to its accurate depth estimation any physical damage, regarding the object under consideration, can be detected which is a major contribution towards an automated quality inspection system.