摘要
本文利用数字图像离散余弦变换系数的广义高斯分布模型,在图像嵌入水印后峰值信噪比(peak signal to noise ratio,PSNR)一致的条件下,以检测概率为衡量指标,理论分析了数字图像加性和乘性算法在未受到攻击时的性能,乘性算法的性能随形状参数的变大而提高,而加性算法的相关检测器性能与形状参数无关。攻击实验分析表明,除了抗滤波攻击性能较差外,加性算法在其他攻击下表现出较强的鲁棒性。
This paper utilizes the generalized Gaussian distribution to model the discrete cosine transform （DCT） coefficients of image. With the same peak signal to noise ratio （PSNR） of watermarked images, the unattacked performances of both multiplicative algorithm and additive algorithm are theoretically ana- lyzed. It is shown that the detection probability of the multiplicative algorithm increases upon the shape parameter of the generalized Gaussian distribution, while the performance of the linear correlation detector for the additive algorithm is irrelevant to this distributed parameter. Besides the attack of filtering, the attacked experiments show that the additive algorithm is preferable to the multiplicative algorithm.
出处
《青岛大学学报（工程技术版）》
CAS
2012年第2期6-10,共5页
Journal of Qingdao University(Engineering & Technology Edition)
基金
山东省自然科学基金项目资助(ZR2010FM006)
关键词
数字水印
广义高斯分布
峰值信噪比
检测概率
digital watermarking
generalized Gaussian distribution
peak signal to noise ratio
detectionprobability
