A self recoverable dual watermarking scheme for copyright protection and integrity verification

Abstract

Dual watermarking implies embedding of robust as well as fragile watermarks into the same cover image. It facilitates integration of copyright protection and integrity verification into the same scheme. However, most of such existing state of art approaches either lacked the feature of tamper detection and original content recovery or provided an approximation using coarser block level approach. The proposed self recoverable dual watermarking scheme integrates all the aforementioned functionalities of copyright protection, tamper detection and recovery into one scheme. The scheme is independent of the order of embedding of robust and fragile watermarks as these are embedded in different regions of the cover image. It performs tamper detection and recovery, both at the pixel level. The scheme obtains recovery information for each 2×2 image block in just eight bits which are further encoded to only four bits via mapping table. This reduction in recovery bits allows efficient embedding of copyright information which is tested against comprehensive set of attacks. The scheme is found to be robust against noises, filtering, histogram equalization, rotation, jpeg compression, motion blur etc. Besides the normalized cross correlation value, the evaluation of the extracted copyright information is also being done using various objective error metrics based on mutual relation between pixels, their values and locations respectively. The imperceptibility and visual quality of the watermarked as well as recovered image is found to be satisfactorily high. Three major categories of images: natural, texture as well as satellite have been tested in the proposed scheme. Even minute alterations can be chalked out as the detection accuracy rate has been enumerated on pixel basis. The scheme can tolerate tampering ratios upto 50 percent though the visual quality of the recovered image deteriorates with increasing tampering ratio. Comparative results based on normalized cross correlation, probability of false acceptance, probability of false rejection and peak signal to noise ratio metrics validate the efficacy of the proposed scheme over other existing state of art approaches.