Signature forgery still represents a great challenge to financial institutions, which makes accurate signature verification inevitable. On the other hand, computer technology and information processing areas witness remarkable qualitative improvements associated with significant costs reduction. This boosted the usage of machine vision techniques. In this research, an intensive work was carried out on offline signatures to establish a system for verifying them using their digital images. Signature morphological structure was utilized to explore characteristics associated with different signatures. Signature verification algorithms were developed using binary images of signatures employing two different verification approaches, one was based on statistical techniques, while the other was based on neural networks (NN) techniques. A signature database was built by collecting 840 signatures from 66 volunteers, and was used for training the statistical and NN classifiers and subsequently for testing purposes. Research results indicated that the statistical classifiers' outcomes were highly satisfactory whereas the NN classifiers' outcomes were not of the same quality.
