Perceptual hashing for hardcopy document authentication using morphological segmentation

Semi-fragile authentication of hardcopy documents is a technique designed to detect any visually signi cant alteration in a document, while ignoring incidental alterations, like distortions resulting from print-scan operations, photocopies, rotations, scalings, translations and minor stains on the paper. It is meant to substitute the use of notarial authenticated photocopies. However, to our knowledge, there is still no functional authentication system for printed documents, only for documents in digital form [1]. A semi-fragile authentication system is composed of three sub-components: perceptual hashing, cryptography and data hiding. This work is concerned with the rst sub-component. The perceptual image hashing h(A) of an image A is a value that identi es A. It is also called robust visual hashing or media hashing [2], [3], [4]. Moreover, given two images A and B, the distance D[h(A), h(B)] between the hashings must be somehow proportional to the perceptual visual di erence of the images A and B. To our knowledge, no perceptual hashing has been proposed for document authentication and perceptual hashings for continuous-tone images cannot be directly applied to authenticate documents. Behera et al. [5] have proposed a perceptual hashing that uses low resolution image of the document as the input data for document retrieval. However, it cannot be used for document authentication, because authentication must detect even the alteration of a single character and must use high-resolution images. This on-going work intend to propose a perceptual hashing for the document authentication.