New Local Difference Feature for Off-Line Handwritten Signature Verification

In this work, we propose a new textural feature for solving offline handwritten signature verification. The proposed feature that is called Local Difference Feature (LDF) is a LBP-like texture descriptor. LDF calculates differences between a central pixel and eight neighbors taken on a specific neighborhood radius. Differences are coded into binary thresholded values before evaluating the histogram of codes. The verification step is achieved by SVM classifier trained on genuine signatures. Furthermore, the test stage is performed on both genuine signatures and skilled forgeries. Exepriments are conducted on GPDS and CEDAR datasets. Results obtained showed that LDF outperforms classical features such as Histogram of oriented gradients and local binary patterns. Also, the comparison with the state of the art highlights the LDF robustness for handwritten signature characterization.