Source Printer Classification Using Printer Specific Local Texture Descriptor

摘要: The knowledge of the source printer can help in printed text document authentication, copyright ownership, and provide important clues about author a fraudulent along with his/her potential means motives. development automated systems for classifying documents based on their printer, using image processing techniques, is gaining lot attention multimedia forensics. Currently, state-of-the-art require that font letters present test unknown origin must be available those used training classifier. In this paper, we attempt to take first step toward overcoming limitation. Specifically, introduce novel specific local texture descriptor. highlight our technique use encoding regrouping strategy small linear-shaped structures composed pixels having similar intensity gradient. results experiments performed two separate datasets show that: 1) publicly dataset, proposed method outperforms algorithms characters same reduces confusion between printers brand model another dataset four different fonts, methods cross experiments.