摘要
为了适应未来精密数字水印的发展,打击侵权盗版行为,在Mallat小波分解的框架之下,利用树结构分层级、多尺度的特性,基于高频带的显著系数,提出了一种全尺度、高信息嵌入量的自适应水印方案,增加了嵌入的信息量和水印鲁棒性。分别在JPEG有损压缩、模糊、增强、几何攻击(裁减和仿射变换)的情况下,根据算法中的嵌入位置和加权因子,并利用全尺度信息进行了水印提取,并用峰值信噪比(PSNR)和相关系数进行嵌入前后图像和水印的比较。该方法显著优势在于实现全尺度的嵌入,显著增加嵌入信息量,更好地折中透明度和鲁棒性。实验结果表明:算法效果明显优于同类其他方法;信息嵌入量大,嵌入的水印对于人类视觉系统(HVS)透明效果好;提取的水印清晰完整,对于各种攻击具有好的鲁棒性。
A scheme of embedding high capability digital watermark in images was developed using a multiple scales method to follow the development of high capability watermark and counterattack tortuous crimes. Under the framework of Mallat wavelet decomposition and based on high frequency significant coefficients, the algorithm realizes full-scale embedding and simultaneously increases the embedding bits and image robustness by taking advantages of the multi-resolutional characteristics of the wavelet tree. The extracting scheme detects watermark by embedding positions and weighting factor associated with the full-scale information for conditions of JPEG loss compression, blurring, sharpening and geometric distortions (i. e., cropping and affine transform). Comparisons between the original and processed images and watermarks by means of PSNR and correlation coefficients show that this algorithm outstands other methods in full-scale embedding, observably increased embedding bits, and excellent trade-off between transparency and robustness. Experimental results also show the large embedded information content, the imperceptibility for the human vision system, neat and intact extracted watermarks, and the robustness against various attacks.
出处
《清华大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2009年第5期749-754,共6页
Journal of Tsinghua University(Science and Technology)
基金
国家自然科学基金资助项目(40704019
40674061)
清华大学基础研究基金资助项目(JC2007030)
国家"九七三"重点基础研究项目(2007CB209505)
关键词
图像信号处理
数字水印
小波树
显著系数
全尺度
自适应
image signal processing
digital watermark
wavelet tree
significant coefficient
full-scale
adaptive