Abstract
Recently, several semi-fragile watermarking approaches with the additional capability of image recovery have been proposed. However, the security, robustness, and image recovery aspect of these approaches have certain shortcomings. In this paper, a novel semi-fragile watermarking framework using integer transform based information embedding and extraction is proposed, which allows accurate authentication and recovery of the image. It is based on integer wavelet transform with improved security against collage attack, enhanced robustness, and capability of producing better quality recovered image. Security is enhanced by correlating the to-be-embedded watermark with the approximation sub-band of wavelet transform. Similarly, no unprotected area is left for attacks on the image, either in spatial or transform domain. Robustness is enhanced by using the idea of embedding in largest coefficient inside a group and correlating it with the quantized version of the mean of the group. In particular, the recovery approach is improved by introducing lossless compression and BCH coding of the integer DCT based low-pass version of the cover image itself. Alteration sensitivity is improved compared to traditional block-based approaches and thus the proposed approach can concisely determine the regions where the integrity verification fails. Experimental comparisons with existing approaches validate the usefulness of the proposed multiple semi-fragile watermarking approach. (C) 2010 Elsevier Inc. All rights reserved.