自适应灰度级数字水印技术

目前的大部分水印技术是侧重于嵌入一维或二维伪随机序列，由一这类水印固有的缺点如信息量少，不具备可视性等，具有明显的说明意义及较好可视性的水印图像正得越来越多的关注，但是这类水印数据量 ，限制了其应用，该文提出了一种新颖的基于小波域的自适应灰度级数字水印技术，包括水印图像预处理和自适应嵌入策略两部分，实验结果证明该方法较好地克服了灰度水印数据量大的缺点，并且有效地提高了水印的稳定性。

Adaptive Gray-Level Digital Watermarking Technique

Most of digital watermarks used now are one-dimensional or two-dimensional pseudorandom sequences. These watermarks have some shortages such as with less information and visually meaningless. So watermark images with more information and good perception are more attractive. However, this type of watermark has too much data payload, which limits its applications. In this paper, we use grayscale image instead of binary image as the watermark image, which is visually meaningful but with more difficulty in its implementation. To deal with the huge amount of data introduced, image preprocessing is adopted before embedding process, in which the watermark image is first decomposed using two layer wavelet transformation and a simple and effective affine technique is used to further decompose the coefficients into bit sequences. By controlling the affine parameter, watermark sequence with less data payload can be obtained. The coarse coefficients and detail coefficients are processed separately with different affine parameter so that the quality of the recovered watermark image can be preserved. To balance the relationship between watermark robustness and imperceptibility, the watermark sequences are embedded into wavelet coefficients of original image with properly selected power control factors. By analyzing the contribution of the bits in different sequences and different position, two conclusions are obtained: (1) coarse watermark sequence is much more important for the reconstruction of watermark image than detail watermark sequence; (2) the significance of different bits in one coefficient is also different. Based on them, we propose a novel adaptive embedding strategy, in which coarse and detail watermark sequences are embedded into coarse and detail coefficients of original image, respectively, and each bit is embedded by its significance order. Experimental results show that the proposed technology can reduce the data payload of grayscale watermark greatly. The watermarked image has good perceptual quality. The recovered watermark image has good quality even under JPEG compression with a quality factor of 30. For other distortions such as Additive White Gaussian Noise (AWGN), image resizing and media filtering, the quality of recovered watermark image is also satisfying. Comparison with common algorithm is shown to demonstrate the effective of adaptive embedding strategy.