Watershed Segmentation for Printed Source Classification

It’s acknowledgeable that different printers would produce different light and dark patterns that are not consistent with the direction in which the paper moves through the printer. Hence, the purpose of this study is to ascertain the differences in the printed characteristics of each printer. Moreover, the experimental procedure includes Watershed segmentation, image processing, classification using a neural network method, and performance evaluation using the Area Under the Curve (AUC) metric. The study uses an Arabic letter collection, with a particular emphasis on the letter "alif." The findings show the efficiency of various segmentation strategies and emphasize the need of selecting appropriate techniques for different printer types. The Dams segmentation approach consistently produces high AUC values ranging from 0.96 to 0.98 across all printer models, demonstrating better effectiveness in detecting counterfeit papers. This research has contributed to advancing the area of document authentication by directing the development of strong systems to prevent document forgery and improve security and trust in various applications.