A hierarchical approach to extracting polygons based on perceptual grouping

Presented in this paper is a method of extracting linear components and polygons in images based on perceptual grouping. In order to reduce the search space which occurs in grouping linear primitives, the method introduces a token-sorting algorithm, where the primitives are sorted according to their slopes and the x-y coordinates of their ending points after the tokens are rotated by the angle of the slope. Experimental results demonstrate that the proposed method is more noise-insensitive and more efficient than conventional methods based on the Hough transform in the aspects of storage and computation