Offline signature verification and quality characterization using poset-oriented grid features

This paper proposes a novel grid-based template matching scheme for off-line signature analysis and verification. At the heart of the new method lies the efficient encoding of the signature׳s fine geometric structure by grid templates, appropriately partitioned in subsets. Features represent the detection of ordered transitions using lattice shaped probing structures shaped on 5×5 pixel window binary masks. The verification performance of the method is evaluated on four different signature datasets producing state of the art results. Additionally, quality characterization of genuine signatures by means of complexity, stability and overall complexity-quality is also carried out. It is shown that both complexity and overall complexity measures correlate strongly with the corresponding opinions expressed by four forensic handwriting experts using the Spearman ranking test. Examination of the verification results provides evidence that the probability of correctly classifying a questioned signature is significantly enhanced when the genuine samples of a signer exhibit higher quality.