Shape from Contour Using Symmetries

Kanade [1] proposed a heuristic for interpreting shape from contour. To apply his heuristic, skewed symmetries have to be found. In this paper, an algorithm for finding skewed symmetries in a planar point set is proposed. The method requires a simple rotation and midpoint finding, followed by a Hough transform. A variant of the Hough transform is reported. It is based on interpreting it as a rotation followed by a projection. This variant requires only a one dimensional accumulator array. An implementation is outlined. We then show (theoretically) that the skewed symmetry finding algorithm is as robust as the standard Hough transform (Hough transform for finding lines).