1 简介
近年来，随着计算机视觉领域的发展，雾天图像的清晰化问题逐渐成为该领域的研究热点。目前对于图像去雾技术的研究大体上可以分为两类[1]：基于大气散射物理模型的图像去雾方法和基于图像增强的去雾方法。基于大气散射物理模型的图像去雾方法从造成图像退化的具体原因角度分析，侧重于考虑大气散射作用对成像过程的影响，对雾天图像的退化过程进行数学建
模，求解模型参数，再通过图像退化的逆过程来恢复原来的真实图像。基于图像增强的去雾方法不考虑图像退化的原因，而是有选择性的突出图像中感兴趣的特征，减弱不需要的特征。一般通过增强图像的对比度和校正图像的颜色以达到图像去雾的目的，从而使处理后的图像更适合人类的视觉特性或计算机识别系统。基于图像增强的去雾方法以其方法简单、有效而得到更为广泛的应用。
2 部分代码
clear all;clc;warning off; close all;foldnumber = 2;linenumber = 150;string = strcat('C:\Users\hehesjtu\Desktop\test\',num2str(foldnumber),'\');Files = dir(strcat(string,'*.bmp')); LengthFiles = length(Files);t = [1:256];for i = 1:LengthFiles string_GT = Files(i).name(end-5:end-4); if (strcmp(string_GT,'GT')) ygt = getline(imread(strcat(string,Files(i).name)),linenumber); name = strcat(Files(i).name(1:end-6)); break; endendpicture = imread(strcat(string,name,'noisy.bmp'));block = ones(5,256,3);block(:,:,1)= block(:,:,1)*255;block(:,:,2) = 0; block(:,:,3) = 0;picture(linenumber-2:linenumber+2,:,:) = block;ynoisy = getline(imread(strcat(string,name,'noisy.bmp')),linenumber);yNLM = getline(imread(strcat(string,name,'NLM.bmp')),linenumber);yBM3D = getline(imread(strcat(string,name,'BM3D.bmp')),linenumber);yDDID = getline(imread(strcat(string,name,'DDID.bmp')),linenumber);yEPLL = getline(imread(strcat(string,name,'EPLL.bmp')),linenumber);yPGPD = getline(imread(strcat(string,name,'PGPD.bmp')),linenumber);yWNNM = getline(imread(strcat(string,name,'WNNM.bmp')),linenumber);yMCWNNM = getline(imread(strcat(string,name,'MCWNNM.bmp')),linenumber);yDnCNN = getline(imread(strcat(string,name,'DnCNN.bmp')),linenumber);yFFDNet = getline(imread(strcat(string,name,'FFDNet.bmp')),linenumber);yOur = getline(imread(strcat(string,name,'Our.bmp')),linenumber);subplot(3,4,1)imshow(picture);subplot(3,4,2)plot(t,ygt,'k','linewidth',2);hold on;plot(t,ynoisy,'r','linewidth',1.5);axis([1 256 50 150]);legend('y_{gt}','y');subplot(3,4,3)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yNLM,'g','linewidth',1.5);axis([1 256 50 150]);legend('y_{gt}','NLM');subplot(3,4,4)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yBM3D,'g','linewidth',1.5);axis([1 256 80 150]);legend('y_{gt}','BM3D');subplot(3,4,5)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yDDID,'g','linewidth',1.5);axis([1 256 80 150]);legend('y_{gt}','DDID');subplot(3,4,6)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yEPLL,'g','linewidth',1.5);axis([1 256 80 150]);legend('y_{gt}','EPLL');subplot(3,4,7)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yPGPD,'g','linewidth',1.5);axis([1 256 80 150]);legend('y_{gt}','PGPD');subplot(3,4,8)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yWNNM,'g','linewidth',1.5);axis([1 256 80 150]);legend('y_{gt}','WNNM');subplot(3,4,9)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yMCWNNM,'g','linewidth',1.5);axis([1 256 80 150]);legend('y_{gt}','MCWNNM');subplot(3,4,10)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yDnCNN,'g','linewidth',1.5);axis([1 256 80 150]);legend('y_{gt}','DnCNN');subplot(3,4,11)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yFFDNet,'g','linewidth',1.5);axis([1 256 80 150]);legend('y_{gt}','FFDNet');subplot(3,4,12)plot(t,ygt,'k','linewidth',2);hold on;plot(t,yOur,'g','linewidth',1.5);axis([1 256 80 150]);legend('y_{gt}','Ours');
3 仿真结果
4 参考文献
[1]吴丽丽. 基于自适应小波与三边滤波的超声图像去噪算法[D]. 浙江工业大学, 2016.
