Offline Signature Verification Using Rotated Local Binary Pattern (RLBP)

Abstract

Handwritten signature verification is an important technique for many financial, commercial, and forensic applications. Signature verification is considered critical in machine learning and pattern recognition, while a significant work is being done to eliminate the ambiguity involved in the manual authentication process. The signature verification can include two kinds, depending on the input format: (1) online and (2) offline. The capture of dynamic image requires an electronic device with a stylus, which mainly records dynamic information while signing. On the other hand, a scanner or any other form of image devices normally capture the offline signature that produces two-dimensional image. There are different approaches for identification of signatures with many areas of research. In this paper, we propose Rotated Local Binary Pattern (RLBP) for feature extraction for the static signature recognition to identify forgeries. The experiments have been conducted using UTSIG dataset. Performance evaluation is done on the basis of accuracy, computation time, sensitivity, specificity, recall and f1- score.