Off-line Persian Signature Verification using Wavelet-based Fractal Dimension and One-class Gaussian Process

Abstract

This paper presents a novel individual verification approach through off-line handwriting signatures. In our proposed technique, images are decomposed into an series of wavelet sub-bands at some specific levels. In order to generate feature vectors, quantization based on Multi-Resolution Box-Counting (MRBC) fractal dimension algorithm is applied. In the training and verification phase, we rely on the One-Class Gaussian Process (OC-GP) priors. The optimal decision threshold is selected from False Accept Rate (FAR) and False Reject Rate (FRR) curves. The presented technique was extensively tested on two available Persian databases (PHBC and UTSig). Experimental results showed very promising results when compared to state-of-the-art techniques. � 2018 IEEE.