Manuscript received April 18, 2025; revised May 16, 2025; accepted June 19, 2025; published November 25, 2025.

Abstract—Ensuring the security of financial transactions and protecting official signatures on financial documents have always been a priority for individuals and international banks. Bank checks play a crucial role in global financial exchanges, with a vast number processed daily. However, signature forgery poses significant risks, leading to identity theft, financial losses, and reputational damage for both financial institutions and customers. Fraudulent transactions impact a bank’s financial stability and erode consumer trust. Therefore, there is an urgent need for innovative solutions to enhance security and prevent signature-related fraud. This study proposes an automated signature verification system for bank checks, leveraging advancements in image processing and deep learning. Our approach begins with the acquisition of a customer’s handwritten signature from check leaves. The main new idea in our method is using the Canny edge detection filter in the layers of the convolutional network instead of the usual Convolutional Neural Network (CNN) filters. This hybrid approach enhances feature extraction by focusing on critical signature contours, improving classification accuracy. Experimental results show that our CNN-Canny hybrid model reaches 98.8% accuracy, which is better than the 95.1% accuracy of traditional CNN methods. These findings highlight the potential of edgeaware deep learning techniques in strengthening security measures.

Keywords—Canny filter, Convolutional Neural Network (CNN), banking transactions, fake signature, signature detection, deep learning, CNN-Canny

Cite: Noor Fadel*, Bayadir A. Al-Himyari, Hiba Al-Khafaji, and Safaa S. Mohammed, "CNN-Canny-Based Signature Scrutiny Verification Techniques in Banking Applications," Journal of Image and Graphics, Vol. 13, No. 6, pp. 630-636, 2025.

Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.