New Method to Detect Text Fabrication in Scanned Documents

Abstract

In our life, we rely on the documents and papers almost in all transaction areas, and with the importance of these documents, there is a new crime appeared in the society called, documents forgery, and it is one of the most serious crimes in societies. Many researchers try to find ways to detect that forgery in order to prevent these danger crimes so they use some methods and approaches to enhance the performance of detection, but with variance performance between their results depend on the difference between their methods. This research introduces a new method, which has two stages and five processes to detect the forgery on scanned document. The first three processes using to preprocessing the documents to them suitable for the next two stages, the fourth process is included in stage one, which using to Extract Max Frequency Intensity feature (EMFI) from the documents pixels, and the fifth process is included in stage two, which using to Extract Edge Gradient (EEG) to find the variance between the real text and the fabricated text. The final process is the serial combination, coloring and locating the suspected pixels. Moreover, we built our own dataset by fabricating some official documents from Islamic University in Gaza (IUG) to test and evaluate the performance of our method, and we measured the system performance for the serial combining between EMFI and EEG methods, by using the recall and precision from the confusion matrix