Scanner model classification with characteristic brightness variations

Analog documents and scanned digitized files are now considered equivalent in legal contexts, and the widespread supply of multi-functional printers has led to a surge in the use of scanned documents. With image editing tools, there has been more cases of forgery involving scanned files. This has highlighted the importance of integrity and authenticity verification of scanned documents submitted as court evidence. Extensive studies have been conducted on source scanner identification and detection of alteration in scanned documents. Past research usually relied on machine learning with Support Vector Machine (SVM) and Convolutional Neural Network (CNN), and was focused more on images rather than text documents. Brightness variations are produced depending on the repetitive arrangement and relative intensity of light sources, and such patterns can be clearly observed in scanned images by the Charged Coupled Device (CCD) type flatbed scanner. The separate image module of the Contact Image Sensor (CIS) also leads to characteristic brightness variations. To extract and enhance these brightness variations, image processing techniques such as separating color channel and adjusting gradation and contrast are applied. The proposed method was tested on five scanner models, and the results confirmed that each scanner had unique brightness variations. This study is the first to extract brightness variations as a unique characteristic of each scanner model and recognize the potential of brightness variations in source identification and manipulation detection. A major advantage is that brightness variations are physical, robust, and visible. The research will be expanded with multicolor documents, counterfeit documents, and text-independent detection.