Image-adaptive Watermarking Using The Improved Signal to Noise Ratio

Conference Paper · November 2006with7 Reads
DOI: 10.1007/978-3-540-74377-4_64 · Source: DBLP
Conference: Computational Intelligence and Security, International Conference, CIS 2006, Guangzhou, China, November 3-6, 2006, Revised Selected Papers


    The two conflicting requirements of watermark invisibility and robustness are both required in most applications. The solution
    is to utilize a suitable perceptual quality metric (PQM) for watermarking correctly. This paper develops a new quality metric,
    the improved signal to noise ratio (iSNR). The improvement is done in the following two aspects: 1) SNR manifests much better
    performance in an image block of small size than in a whole image; 2) the average luminance and gradient information are added
    into SNR. Next, we propose a new adaptive watermarking framework based on the localized quality evaluation, which divides
    the cover data into nonoverlapping blocks and assigns an independent distortion constraint to each block to control the quality
    of it. In comparison with ones based on the global quality evaluation, the new one exploits the localized signal characteristics
    sufficiently while guaranteeing the localized watermark invisibility. Then, a specific implementation of the above framework
    is developed for image applying iSNR as the quality metric in the sense of maximizing the detection value. Experimental results
    demonstrate that the proposed watermarking performs very well both in robustness and invisibility.