Data hiding for authentication of binary images
Huijuan Yang · 2008
The advancement in digital technology and multimedia has made it easy to duplicate, manipulate and illegally distribute the multimedia content, which brings serious concerns in copyright protection and authentication.In addition, it is easy to forge an image with the powerful image editing software and digital cameras.In this thesis, we propose data hiding techniques that operate on different level (e.g., object and pixel) and on different domain (e.g., spatial and transform) for the purpose of binary images authentication and tampering localization.Issues in capacity, visual quality, system security, computational complexity and accuracy in tampering localization are addressed.The validity of the methods is demonstrated by both quantitative results and theoretical analysis.We firstly investigate data hiding for text images in object level by integrating the inter character and word spaces.A new concept of maximum inter character space is introduced such that the relationship between the inter character and word spaces can be established.The security level of the data hiding system has been enhanced by randomly shifting the characters as well as by embedding the authenticator watermark.This method can be applied to a wide variety of text documents authentication, especially for the applications that do not allow the distortions introduced by flipping pixels in binary images.How to identify the suitable locations to carry the information data is a key issue for data hiding techniques operating in pixel level.We propose a "Connectivity-Preserving" criterion to assess the "flippability" of a pixel to preserve the connectivity of pixels in a ATTENTION: The Singapore