基于聚类分析的小波域数字水印算法

Digital watermarking algorithm in the wavelet domain based on cluster analysis

所谓数字图像水印是指向数字图像中加入秘密信息如随机序列、标识性图像等,并且在需要的时候能够被正确地提取或检测出来以达到版权保护的目的.该文提出了一种小波域内基于聚类分析嵌入水印的算法.该算法将原始图像分解到小波域,通过聚类将小波系数分成若干类,根据分类的性质选取部分小波系数嵌入水印.提取水印时不需要原始图像,是一种盲水印算法.实验结果表明,该算法对JPEG压缩、剪切、中值滤波、加噪等具有较好的鲁棒性.

Digital image watermarking is a process in which secret information such as random sequences, identified pictures , and so on are embedded into digital images to be extracted or detected intact when needed, to prove the copyright ownership of digital images. In this paper, a watermarking algorithm based on cluster analysis in the wavelet domain is proposed. In this method, the original image is decomposed into the wavelet domain, and the wavelet coefficients are classified into several classes by clustering. Afterwards some coefficients are chosen to embed as part of a watermark according to the property of every class. This algorithm does not compare the test image with the original image when evaluating the watermark, so it is a blind watermarking algorithm. Experimental results show that it is a highly robust algorithm against JPEG compression and cropping, median filtering and the addition of noise.