New Textural Features for Handwritten Signature Image Verification

In this paper, a new textural feature for solving offline handwritten signature verification is proposed. A new textural features method is developed by combining a Local Binary Patterns (LBP) method and a Gradient Quantization Angle (GQA) method. This proposed method is called Local Binary Patterns with Gradient Quantization Angle (LBPGQA), as developed by heuristic method to improve the precision of verification the offline signature image. The hypothesis for this study is to classify the distinctive handwritten signature individually with the actual signature angle and refraction for enhancing the signature fraud detection. The verification step is achieved by Artificial Neural Network (ANN) classifier and trained on genuine signatures. Furthermore, the test stage is performed on genuine signatures and skilled forgeries. The experiments are conducted on CEDAR datasets. The experimental results show that in the LBPGQA method outperforms classical features such as Histogram of oriented gradients and local binary patterns. Conclusively, this proposed method can verify the individual and distinctive handwritten signature and help to protect the signature fraud by skilled forgeries.