Feature Based Robust Digital Image Watermarking Scheme

A robust digital image watermarking scheme that combines image feature extraction and image normalization is proposed. The goal is to resist both geometric distortion and signal processing attacks. We adopt a feature extraction method called Mexican Hat wavelet scale interaction. The extracted feature points can survive a variety of attacks and be used as reference points for both watermark embedding and detection. The normalized image of an image (object) is nearly invariant with respect to rotations. As a result, the watermark detection task can be much simplified when it is applied to the normalized image. However, because image normalization is sensitive to image local variation, we apply image normalization to nonoverlapped image disks separately. The disks are centered at the extracted feature points. Several copies of a 16-bit watermark sequence are embedded in the original image to improve the robustness of watermarks. Simulation results show that our scheme can survive low-quality JPEG compression, color reduction, sharpening, Gaussian filtering, median filtering, row or column removal, shearing, rotation, local warping, cropping, and linear geometric transformations.

MANY digital watermarking schemes have been proposed for copyright protection recently due to the rapid growth of multimedia data distribution. On the other hand, attacks have been developed to destroy watermarks. These attacks on watermarks can roughly be classified as geometric distortions and noise-like signal processing. Geometric distortions are difficult to tackle. They can induce synchronization errors between the extracted watermark and the original watermark during the detection process, even though the watermark still exists in the watermarked image. Nowadays, several approaches that counterattack geometric distortions have been developed. These schemes can be roughly divided into invariant transform domain-based, moment-based, and feature extraction-based algorithms. Watermarks embedded in invariant-transform domains generally maintain synchronization under rotation, scaling, and translation. Examples of these transforms are log-polar mapping of DFT and fractal transform coefficients . A structured template may be embedded in the DFT domain to assist watermark synchronization during the detection process [3], [4]. The template should be invisible and have low interference with the previously embedded watermarks. A fixed structured template may be identified and destroyed easily. Watermarks embedded in the DFT domain are sensitive to other types of geometric transformation such as local warping. There is an accuracy problem associated with log-polar mapping of DFT since the inverse transformation requires image interpolation. The watermark detection process is similar to the pattern recognition process in computer vision, but the original images may not be available to the watermark detector. Moments of objects have been widely used in pattern recognition. Higher order moments are more sensitive to noise, and some normalization schemes have been designed to tolerate noise . A watermarking system employing image normalization with respect to orientation and scaling is proposed in . If the image normalization process is applied to the entire image, it would be sensitive to cropping and local region distortion. Another moment-based watermarking scheme hides watermarks by modifying image content iteratively to produce the mean value of several invariant moments in a predefined range. The watermark detector verifies the presence of the watermark by checking the mean value of these moments. This scheme can resist orthogonal transformations and general affine transformation, but it is sensitive to cropping and aspect ratio changes

Free download research paper