Analysis on the Robustness SVD-Based Watermarking Algorithms

This paper deals with the testing of the robustness of two algorithms for embedding and extraction of digital watermark into a monochromatic image, based on SVD transformation. The first algorithm was proposed by Liu and Tan [1], and the second, named Technique 1, was proposed by Mohammad et al. [3]. In the first part of the paper, the influence of the watermark contents on the the image for both algorithms was analyzed. Thus, the optimal scale factor of embedding watermark was determined. Then, the images with the watermark were tested under the conditions where various deformations were applied. The robustness was analyzed as well on the basis of objective and subjective parameters. In the second part of paper, impulse noise with percentage of 10-50% was added to the watermark images. Then, an efficient algorithm for impulse noise elimination was applied. The watermark was extracted from the reconstructed images and its quality was analyzed on the basis of subjective and objective parameters. PSNR and correlation coefficient rcorr were used as objective parameters. The visibility of the extracted watermark was used as a subjective parameter. On the basis of the results, an estimation of the robustness of the tested algorithms was made. Keywords-Image processing; digital watermarking; SVD transformation; impulse noise;