改进的局部最小像素先验遥感图像盲复原算法  

Improved local minimum pixel prior for blind restoration algorithm of remote sensing images

作　　者：朱兵[1] 王晨 朱福珍[2] 王曼威 ZHU Bing;WANG Chen;ZHU Fuzhen;WANG Manwei(School of Electronics and Information Engineering,Harbin Institute of Technology,Harbin 150001;College of Electrical Engineering,Heilongjiang University,Harbin 150080)

机构地区：[1]哈尔滨工业大学电子与信息工程学院,哈尔滨150001 [2]黑龙江大学电子工程学院,哈尔滨150080

出　　处：《高技术通讯》2024年第2期123-131,共9页Chinese High Technology Letters

基　　金：国家自然科学基金(61601174)资助项目。

摘　　要：为了解决遥感图像盲复原时模糊核估计不准确、复原图像存在振铃效应的问题,提出改进的局部最小像素先验遥感图像盲复原算法。该算法首先引入极端通道先验与局部最小像素先验结合,对图像的强度进行更好的约束,有利于得到更好的潜在清晰图像;然后采用基于梯度的方法估计模糊核,模糊核估计与中间潜在清晰图像估计交替迭代进行,获得较为理想的模糊核;最后引入联合双边滤波器,采用改进的拉普拉斯与正则化图像复原算法抑制图像复原的振铃效应。实验结果表明,本文方法对遥感图像复原效果较好,恢复的图像边缘清晰,振铃伪影得到抑制且模糊核较为理想;客观评价指标峰值信噪比(PSNR)较前沿复原算法平均提高约1.40 dB,结构相似度(SSIM)平均提高约0.02。In order to solve the problem of inaccurate fuzzy kernel estimation and ringing effect in restored images during blind restoration of remote sensing images,an improved local minimum pixel a priori remote sensing image blind restoration algorithm is proposed.The algorithm introduces the combination of extreme channel a priori and local minimum pixel a priori to better constrain the intensity of the image,which is conducive to obtaining a better potentially clear image;then the gradient-based method is used to estimate the fuzzy kernel,and the fuzzy kernel estimation is carried out alternatively and iteratively with the estimation of the intermediate potentially clear image to obtain a more desirable fuzzy kernel.Finally,an improved Laplace and regularized image restoration algorithm is used to input the resulting fuzzy kernel.That is,a joint bilateral filter is introduced to suppress the ringing effect of image restoration.The experimental results show that the method in this paper has a good effect on remote sensing image restoration,the restored image has clear edges,the ringing artifacts are suppressed and the fuzzy kernel is more ideal.The objective evaluation index the peak signal to noise ratio(PSNR) is improved by about 1.40 dB and the structural similarity(SSIM) is increased by about 0.02 on average compared with the cutting-edge restoration algorithm.

关 键 词：图像盲复原 通道先验 局部最小像素先验 联合双边滤波器 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]