Abstract: This paper presents a blind watermark algorithm of multi-wavelet domain based on neighborhood and FCM. The watermark embedding texture sub-blocks are obtained adaptively on the multi-wavelet domain middle-frequency coefficients by applying FCM based on the human visual system. A binary image is embedded to texture sub-blocks, this algorithm can get a watermarked image with perceptual invisibility and strong robustness. In order to improve the security and robustness, the watermark is scrambled before embedded, and is repeated when embedded. Experimental results show that the algorithm is strongly robust for the common image processing techniques such as JPEG compression, cropping, sharpening salt and pepper noise, filtering and so on.

Key words: multi-wavelet transform, fuzzy clustering, watermark

摘要： 提出一种基于邻域与FCM的多小波域盲水印算法。该算法根据人类视觉系统的掩蔽性对多小波域的中频系数进行模糊聚类分 析，自适应地找到适合嵌入水印的纹理边缘子块，并结合子块的局部相关性嵌入二值水印图像。为保证增强水印的安全性和鲁棒性，在嵌入前将水印置乱，嵌入时重复嵌入水印。实验结果表明，该算法对JPEG压缩、剪切、锐化及边缘锐化、椒盐噪声、滤波等常规攻击有较强的鲁棒性。

关键词: 多小波变换, 模糊聚类, 水印

CLC Number: 

• TP309