Off-line signature verification based on chain code histogram and Support Vector Machine

In this paper, we present an approach based on chain code histogram features enhanced through Laplacian of Gaussian filter for off-line signature verification. In the proposed approach, the four-directional chain code histogram of each grid on the contour of the signature image is extracted. The Laplacian of Gaussian filter is used to enhance the extracted features of each signature sample. Thus, the extracted and enhanced features of all signature samples of the off-line signature dataset constitute the knowledge base. Subsequently, the Support Vector Machine (SVM) classifier is used as the verification tool. The SVM is trained with the randomly selected training sample's features including genuine and random forgeries and tested with the remaining untrained genuine along with the skilled forge sample features to classify the tested/questioned sample as genuine or forge. Similar to the real time scenario, in the proposed approach we have not considered the skilled fore sample to train the classifier. Extensive experimentations have been conducted to exhibit the performance of the proposed approach on the publicly available datasets namely, CEDAR, GPDS-100 and MUKOS, a regional language dataset. The state-of-art off-line signature verification methods are considered for comparative study to justify the feasibility of the proposed approach for off-line signature verification and to reveal its accuracy over the existing approaches.