ROBUST CONTOUR DECOMPOSITION USING A CONSTANT CURVATURE CRITERION

In many computer vision paradigms it is necessary to decompose an extended boundary or contour into simple primitives. Such decompositions are necessary for further structural or syntactic analysis and concise description of the contour shape. We address this problem with particular emphasis on Laplacian-of-Gaussian (LoG) zero crossing contours. We introduce a novel technique for partitioning such contours into constant curvature segments. In doing so, we present a nonlinear "blip" filter matched to the impairment signature of the curvature computation process, an overlapped voting scheme, and a sequential contiguous segment extraction mechanism. This technique is insensitive to reasonable changes in algorithm parameters and robust to noise and minor viewpoint-induced distortions in the contour shape such as those encountered between stereo image pairs, an application of special interest. The results vary smoothly with the data and local perturbations induce only local changes in the result. Robustness and insensitivity are experimentally verified.