A Comparative Study on Image Hashing for Document Authentication

A digital seal is a cryptographically signed 2D barcode printed on a document to verify the document''s authenticity and integrity. In order to secure a printed image against tampering, a compact image hash can be stored inside the barcode. We investigate and experimentally evaluate various methods from the areas of image hashing and face verification w.r.t. its appropriateness for such an image hash that is resistant to a print-scan transformation. Exhaustive experiments with genuine security paper show that compressed local binary patterns (LBPs) and feature extraction by DCT-II perform best in this setting.