Off-line signature verification by local granulometric size distributions

摘要: A fundamental problem in the field of off-line signature verification is lack a representation based on shape descriptors and pertinent features. The main difficulty lies local variability writing trace which closely related to identity human beings. In this paper, we propose new formalism for visual perception. image consists 512/spl times/128 pixels centered grid rectangular retinas are excited by portions signature. Granulometric size distributions used definition an attempt characterize amount signal activity exciting each retina focus attention grid. Experimental evaluation scheme made using database 800 genuine signatures from 20 individuals. Two types classifiers, nearest neighbor threshold classifier, show total error rate below 0.02 percent 1.0 percent, respectively, context random forgeries.