ROBUST WATERMARKING OF COLOR IMAGES USING RST INVARIENT FEATURES

Watermark refers to digital data embedded in cover data with an intent of copyright protection or proof of ownership. In case of image watermarking, the robustness of watermarks to geometric attacks is considered to be the most challenging design requirements for watermarks. Geometric attacks like Rotation, Scaling and Translation (RST attacks) can desynchronize the location of the watermark and hence cause incorrect watermark detection. In this paper, a new color image watermarking scheme for copyright protection is proposed. This scheme utilizes the features of the image which are invariant under rotation, scaling and translation of images. Moments and functions of moments have been extensively employed as invariant global features of images in pattern recognition. The proposed work utilizes Hu's invariant moments for watermark generation. The same can be detected and extracted at the detector to serve for copyright protection. The moments are invariant against image rotation, scaling and translation. Thus, the proposed watermarking scheme is robust against RST attack. Original unmarked image is not required at the detector for watermark detection and extraction. Thus, the proposed scheme provides a way for blind watermarking. The simulation is done using MATLAB and results are compared with existing techniques of invariant feature based watermarking schemes. Comparison of the results are presented which shows a considerable improvement in Peak Signal to Noise Ratio.