摘要
针对现有的数字图像水印算法在应用于医学图像时过于复杂,接收端定位感兴趣区域处理时间长的问题,提出一种依据特征值自动还原感兴趣区域的双水印嵌入算法.使用Harris角点法提取医学图像特征值,根据特征值分布划分感兴趣区域,使用奇异值分解将认证水印嵌入感兴趣区域经过轮廓波分解后的低通子带的最大奇异值中,病人信息嵌入至经过轮廓波分解后的非感兴趣区域中频子带系数中,并选择一定数量的感兴趣区域稳定特征点值嵌入至高频子带中.实验结果显示嵌入水印前后特征提取匹配程度高,篡改探测实验成功率均大于90%,PSNR值大于42db.算法通过提取特征值实现了感兴趣区域自动选择,提高实际应用效率,充分利用轮廓波分解得到多个子带提高图像水印负载能力,与奇异值分解法结合保证水印具有良好的不可见性和抗攻击能力,增强医学图像传输的安全性和完整性.
To improve the problems of digital watermark’s complexity in practical application of medical image and long-time processing on region of interest position,this paper proposed a dual watermarking algorithm with function of ROI automatic orientation. Firstly used Harris’ s method to extract corner points in medical images. and marked ROI based on the distribution of characteristic regions.Then applied contourlet transformation to ROI and RONI and implanted the authentication watermark into the maximum singular value of the lowpass subband. Then embedded patient information into RONI’s middle-frequency coefficients selected a block from RONI and applied Contourlet transformation to get its high frequency sub-band. The values of characteristic points were embedded into RONI’s high-frequency subband using SVD. Experiments showthat characteristic points extracting from the original images and watermark-embedded images are matched. The success rates of detecting tamper are all greater than 90%. PSNR values are greater than42 db. The algorithm realizes ROI automatic orientation and improves the efficiency of practical application. With the full advantage of contourlet decomposition,the capacity of watermark embedding is improved. The combination with singular value decomposition ensures watermarks with good invisibility and attack-resistance.
作者
陈青
宗偲琦
CHEN Qing;ZONG Si-qi(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2019年第1期205-209,共5页
Journal of Chinese Computer Systems
