Approach to the chinese seal registration

Abstract

Registration becomes difficult when the concerned pattern - e.g., a Chinese seal pattern - is noisy and rotated. An approach to Chinese seal registration is presented. For a color image, a self-organization feature map algorithm and a simple color filtering process are applied for seal segmentation. Then, a contouring analysis is performed on the segmented seal to detect its principal orientation. Finally, the registration is achieved by combining the orientation difference and the center region translation for two seals. Experiments on seals with rotations, heavy noise, and different resolutions confirm the feasibility of the proposed approach. © Springer 2008.