A signature verification system usually consists of a learning stage and a testing stage. Always an assumption is made by most previous studies that the variations in orientation and scale of the signatures done by the signer in the two stages should be minimized, otherwise, errors might occur. In this study, we propose a new method to solve the problem. By treating each signature as a point pattern consisting of a set of critical points, the relaxation method proposed by reference (15) is extended first, acid next the best matching pairs are found using the state-space search method like the A* algorithm. According to the matching results, an affine transformation matrix is computed finally. By applying the matrix to the signatures, the variation problem mentioned above is thus solved. Experimental results are given to show the superiority of the method. (C) 1998 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved. [References: 18]
展开▼