Due to extension of computer networks and Internet, digital contents can nowadays be easily exchanged. As a result many copyrighted contents may be shared illegally and this has concerned multimedia owners. To overcome this problem, watermarking schemes are proposed to protect copyright of digital contents such as digital images. These techniques are divided into two groups of spatial and frequency domain methods. A proper watermarking scheme must have high robustness as well as preserving visual quality of the watermarked image. Since that robustness and fidelity are in contrast to each other, having these features at the same time is a major advantage for a robust watermarking method. In this thesis, after discussing watermarking concepts and introducing different types of watermarks, attacks, and several watermarking methods, four watermarking algorithms are proposed. Since proposing a robust watermarking method for digital images is of concern in this thesis, and since frequency domain methods are more robust than spatial domain techniques, all of the proposed watermarking algorithms are designed in the frequency domain. One of the four presented methods is based on wavelet transform. In this algorithm, a HWD based mask is designed to choose proper coefficients of the LL subband for embedding. Human eye is less sensitive to modifications of 2D singularities and since Contourlet transform can extract 2D singularities better than wavelet, hence, the other proposed methods are based on embedding in Contourlet coefficients. To increase robustness, two of the proposed schemes embed the watermark into singular values of Contourlet coefficients. Since proper value of strength factor has an important effect on fidelity and robustness of watermarked images, in this thesis a genetic algorithm is proposed to determine this strength parameter in the embedding phase. The fitness function of genetic algorithm considers both fidelity and robustness of a watermarked image. Three out of four proposed schemes are non-blind methods and need the original image in the extraction phase. But one of the schemes is a blind algorithm and can extract the watermark without having the original image. Since sometimes the original image may not be available at the extraction phase, blindness is a major advantage of this algorithm. To increase robustness of the blind method, embedding is done in DCT coefficients of Contourlet blocks. Furthermore, all the Contourlet based methods benefit from multiple embedding of watermark that increases robustness of proposed techniques. The results of experiments show that proposed schemes have good robustness against usual image processing attacks while preserving fidelity of watermarked images. Keywords: Watermark, watermarking, robustness, fidelity
