THINNING APPROACH FOR NOISY DIGITAL PATTERNS

摘要

Many problems difficult to solve by traditional computer algorithms can be explored by the principles of human visual perception. From the hint of human visual symmetry, a thinning approach for noisy digital patterns is presented in this paper. It includes three main algorithms. Algorithm 1 is designed to find the effective circular range, for each pixel, containing the maximum effective information. Algorithm 2 is designed to compute the symmetry score of the pixel distribution in the circular range and convert the symmetry information into a gray-scale image. Algorithm 3 provides a modified gray-scale thinning algorithm for thinning the gray-scale image to obtain a connected thin line. Experiments show that the thinning results obtained by the proposed method are properly acceptable when SNR greater than or equal to 2.2 dB and the connectedness of the thin line can be preserved. Copyright (C) 1996 Pattern Recognition Society. [References: 31]