Contour Stereo Matching by Correlation in a Constraint Satisfaction Approach

Stereo vision is the recovery of depth of a scene using multiple images with di erent points of view. The major task in stereo vision is establishing correspondence of points between images given a known pinhole-model of the stereo cameras. Two major approaches to matching points have been used widely: feature-based matching and areaor correlationbased matching. It is well known that both approaches have speci c advantages and disadvantages. We describe a method for the matching of contours in a constraint satisfaction approach. Contour matches allow for depth estimation of textured surfaces and objects boundaries to sub-pixel accuracy. These depth estimates may be used for geometric object modeling. We analyze and provide quantative results of the constraint-based stereo matching, and additionally present some examples of objects models obtained with our approach.