Volume : 2, Issue : 12, December - 2013

Signature Verification Using Image Processing Techniques

Amina Khatra

Abstract :

The major problem associated with signature verification is the availability of limited data. As signature data are legally accepted as the authentication means for many financial or other official works, this is difficult to have a sufficient amount of data required to develop a signature verification system. As a result, robust parameter estimation on limited sample sets is still one of the major research issues in this field. One technique to address this problem is to extend the techniques of classical model adaptation for discriminative training. The other challenging problem in offline signature verification is the feature extraction process. Choice of features depends on the style of the signatures and hence different styled–signatures will have different characteristic features. So, it is difficult to develop one general system to classify every style of signatures. Signatures in different scripts may not recognized by a single classifier or even a classification system. It has been observed that most of the researchers have proposed or developed their systems for a limited type of signatures. However achieving an acceptable accuracy in various individual signature styles will make it Handwritten signature recognition can be divided into online (or dynamic) and off–line (or static) recognition. Online recognition refers to a process that the signer uses a special pen called a stylus to create his or her signature, producing the pen locations, speeds and pressures, while off–line recognition just deals with signature images acquired by a scanner or a digital camera. In general, offline signature recognition is a challenging problem. Unlike the on–line signature, where dynamic aspects of the signing action are captured directly as the handwriting trajectory, the dynamic information contained in off–line signature is highly degraded. Handwriting features, such as the handwriting order, writing–speed variation, and skilfulness, need to be recovered from the grey–level pixels.In the statistical approach, each pattern is represented in terms of N features and is viewed as a point in a N–dimensional space. The effectiveness of the representation space (feature set) is determined by how well patterns from different classes can be separated.