Digital Image Watermarking Technique Based On Wavelet Transform And HVS

Digital watermark evolved in recent years is a new means to protect IPR (Intelligence Properties Right) in the field of information security. A digital watermark is information embedded in the host data for the purpose of monitoring and tracking the host data. A new algorithm apply for still images is proposed. The embedding process is completed in DWT domain by exploiting the characteristics of HVS. The watermark consists of a pseudorandom sequence and its strength is modulated according to the local image characteristics. In construct to current schemes proposed so far which either operate on a subband or a block base, the scheme in this thesis is based on pixel. In addition, a sophisticated mechanism derived from DWT coding is introduced to take into account several HVS phenomena such as gray-level sensitivity, noise sensibility around edges and isofrequency and nonisofrequency masking. All these characteristic models are used to serve as the factors of the weighing function. The original image is not needed in watermark detection process. The algorithm is based on the optimum quantization step for each DWT coefficient according to psychovisual considerations which derived from methods in image compression. Some modifications are proposed here to make it suitable to the computation of the maximum visibly tolerable watermark energy. Thanks to the similarity of DWT decomposition to the models of HVS, the algorithm is proved to be robust and invisible.