Digital Watermarking Methods for Authentication, Tamper Localization and Lossless Recovery

This thesis addresses authentication watermarking and compression of digital images under zero-distortion and maximum-per-sample-distortion constraints. In this context, a hierarchical authentication watermark is proposed for securely verifying the source and integrity of digital images and for determining altered image regions when the integrity verification fails. The proposed approach eliminates the security problems associated with previous independent block-based authentication watermarks, while sustaining their superior tamper localization properties. Insertion of the authentication watermark induces a small, bounded, yet permanent loss of image fidelity, which is often undesirable. A novel lossless data embedding method remedies this problem and enables zero-distortion reconstruction of unmarked images upon extraction of the embedded payload. The reconstruction is achieved by compressing portions of the image data that are susceptible to watermarking distortion—such as the least significant image levels— and transmitting the compressed descriptions as a part of the embedded payload. An algorithm that utilizes the remainder of the image data as side-information is proposed for efficient compression of the least significant image levels. This algorithm exploits both spatial and inter-level correlations within the image and achieves excellent compression—consequently, lossless data embedding—performance. The algorithm serves also as the basis for a new level-embedded image compression method, which generates a scalable bit-stream, truncation of which corresponds to bit-plane truncation in the pixel domain. Security and tamper localization capability of the hierarchical scheme are combined with the zero-distortion reconstruction property of the lossless data embedding method in a flexible framework to obtain a lossless authentication watermark. As opposed to earlier lossless authentication methods that required recovery of the original image prior to validation, the new framework allows validation of the watermarked images before the reconstruction step. For verified images, integrity of the reconstructed image is ensured by the uniqueness of the recovery procedure. As a result, the framework offers computational efficiency, improved tamper-localization accuracy, and implementation flexibility.