Reversible fragile watermarking based on pyramidal structure and gradient predicting image

The purpose of this research lies in the development of a reversible fragile watermarking scheme that not only can restore the original image but can also both correctly identify and localize unauthorized manipulations to a protected image. The similarities among the detail components of a pyramid-structure image are exploited to select appropriate embedding areas, and then the cryptographic watermarks are embedded into selected wavelet coefficients using a difference-expansion method. No extra space is required to store the watermark information, watermarks and watermark location map. Experimental results show that the proposed scheme can successfully obtain original image if protected image is unaltered, and unauthorized manipulations can be correctly identified and localized even when protected images undergoes a cropping attack.