灰度极值加权求和图像振铃效应评价算法  

Gray Extreme Weighted Sum Image Ringing Evaluation Algorithm

作　　者：柯铭[1] 张天明 秦爱景 王波 白旭 KE Ming;ZHANG Tian-ming;QIN Ai-jing;WANG Bo;BAI Xu(School of Computer and Communieation Lanzhou University of Technology,Lanzhou 730050,China;School of Measurement-control Technology and Communications Engineering,Harbin University of Science and Technology,Harbin 150080,China)

机构地区：[1]兰州理工大学计算机与通信学院,甘肃兰州730050 [2]哈尔滨理工大学测控技术与通信工程学院,黑龙江哈尔滨150080

出　　处：《哈尔滨理工大学学报》2019年第5期93-100,共8页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(61263047)

摘　　要：在图像盲复原过程中,对模糊核的过估计会产生误差,导致图像振铃效应;现有算法无法对振铃的严重程度进行有效评估,针对这个问题,提出一种振铃效应的评价方法。从振铃效应产生的物理机理出发,分析图像盲复原所致振铃效应的原因和对图像质量的影响;通过提取振铃图像过冲与涟波区域,并对该区域梯度值进行加权统计,定量化表征振铃效应的严重程度。实验结果表明,当模糊核参数σ值超过3时,算法的结果曲线随着参数增大呈单调递增状态,SSIM评价结果变化幅度趋于零。算法可以有效地评价不同复原算法和不同复原参数下的复原图像中的振铃效应。In the process of image blind restoration,the estimation of the fuzzy kernel usually produces errors which lead to image ringing.The existing algorithm cannot effectively evaluate the severity of ringing.To solve this problem,a method for evaluating the ringing effect is proposed.From the physical mechanism of image ringing effect caused by blind restoration of image,the causes of the ringing effect caused by blind restoration of images and the influence on image quality are analyzed.By extracting the overshoot and ripple regions of the ringing image and weighting the gradient values of the region,the severity of the ringing effect is quantified.The experimental results show that when the fuzzy kernel parameter value exceeds 3,the algorithm shows a monotonically increasing state with increasing parameters,The change in SSIM evaluation results tends to be zero.This algorithm can effectively evaluate different restoration algorithms and different restoration parameters.Restore the ringing effect in the image under different restoration parameters.

关 键 词：振铃 盲复原 质量评价 灰度极值加权求和算法 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]