DCT结合模糊推理的鲁棒数字水印算法  

A Robust Digital Image Watermarking Algorithm Based on FIS and DCT

作　　者：赵康[1] 刘帅[2] 张少辉 

机构地区：[1]商丘职业技术学院学生处,466001 [2]商丘职业技术学院汽车建筑工程系,河南商丘476100 [3]周口师范学院网络工程学院,河南周口466001

出　　处：《控制工程》2017年第12期2466-2471,共6页Control Engineering of China

基　　金：河南省科技厅软科学研究计划项目(142400411213);河南省高等学校重点科研项目(15A520118)

摘　　要：针对一般变换域的数字水印算法在数字完整性和鲁棒性方面较差的问题,提出一种基于离散余弦变换和模糊推理的鲁棒数字水印图像算法。首先将植入水印的灰度图分成8×8的区域块,并对每个区域块进行DCT运算,获取人类视觉系统HSV参数;然后利用三种模糊推理系统产生嵌入水印的权重因子,模糊推理系统的输入值则是从HSV模型中获得。最后每一个区域块的中间频率处作为水印的植入位置,通过植入公式来完成水印植入。采用峰值信噪比PSNR,结构相似性度量SSIM和归一化互相关函数NCC评估水印植入效果。从实验数据可以看出,所提算法在大多数情况下具有最高的PSNR、SSIM和NCC值,在隐藏性和抗攻击鲁棒性方面,优于其他优秀算法。As the integrity and robustness of common transform domain watermarking algorithms is not very good, a robust digital image watermarking algorithm based on FIS and discrete cosine transform is proposed. Firstly, the grayscale image of watermarks is divided into blocks of 8 plus 8, and each block is calculated with DCT. The parameters of the human visual system are obtained. Then three fuzzy inference systems are used to generate watermarked weighting factors and inputs of the fuzzy inference system is taken from the HSV model. Finally, the watermarking is embedded into the middle frequency, and the embedding is finished by the embedding formula. In this paper, the peak signal to noise ratio PSNR and structural similarity measure SSIM and normalized cross-correlation function NCC are used to evaluate watermarking results. The experimental results show that the proposed algorithm has the highest PSNR, SSIM and NCC values in most cases. The proposed algorithm is better than other outstanding algorithms in the aspects of hidden and robustness to attack.

关 键 词：信息安全 数字水印 图像攻击防御 离散余弦变换 模糊推理系统 权重因子 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]