Off-Line Arabic Signature Recognition Based On Invariant Moments Properties

The signature recognition process is special case of handwritten recognition. The process includes many complexity, where the signature is a dynamic feature that affected by many operators like psychology, the kind of pen and paper and other operators. In this paper, a number of persons were selected to use their signature as a database for the work. Six signatures were taken from each person through two separated period of time using the same pen and paper. The adopted method consists of three main steps. In the first step, the digital image of the signature transformed into contours. After that the main contours were extracted and the noise was rejected. These extracted contours and their dimensions were measured precisely according to their (x) and (y) axis. Second step is the coding step, where the (Chain Code) method was used to code the extracted contour from the first step, converted them into vectors in which they are very easy to deal with. Using length of the vectors were sorted descending by that can be easily used in comparison process. The third (final) step includes application of the (Invariant Moments) method with these chain vectors and the calculated mean of the output for the five signatures taken for each signer and used it as a eference feature for the signer in the recognition process. The signature recognition process completed using the (Minimum Distance) method as a classifier to identify the personal signature.