Analysis on Segmentation and Registration for Chinese Seal Identification
Faculty of Health, Engineering and Science
School of Computer and Security Science / eAgriculture Research Group
Intelligent seal imprint identification has been studies from diverse perspectives for decades. Even though a number of seal registration methods using digital image processing techniques have been proposed, the difficulty of seal registration remains unresolved especially when the concerned patterns are very noisy and rotated. To overcome the problems, we propose an intelligent segmentation for binarizing the input color seal images in terms of the adaptive thresholds. The FFT (Fast Fourier Transform) is applied to register rectangular and elliptic seals in polar coordinates, and the novel spatial spectrum function called ring-projection is used for circular seal registration. Empirical analysis has demonstrated that the proposed method performs much better than other approaches.