自动色阶算法实现

自动色阶算法用于图像增强，去雾等，其思想去掉像素值最高的比例，去掉像素值最低的一些比例，然后在将图像的其余像素值进行线性映射或者伽马校正至[0, 255]区间。如下图，在图像处理中是不是很熟悉?!

本文主要参考人在旅途的博客，采用matlab语言实现，算法不在详细叙述，在此代码留做备注。http://www.cnblogs.com/Imageshop/archive/2011/11/13/2247614.html。

 

以下是程序代码：

 

function imDst = autolevel(varargin)
[I,lowCut,highCut] =parse_inputs(varargin{:});
[hei,wid,~] = size(I);
 
PixelAmount = wid * hei;
if size(I,3)==3
 
   [HistRed,~]  = imhist(I(:,:,1));
   [HistGreen,~] = imhist(I(:,:,2));
   [HistBlue,~] = imhist(I(:,:,3));
 
   
 
   CumRed = cumsum(HistRed);
   CumGreen = cumsum(HistGreen);
   CumBlue = cumsum(HistBlue);
 
   
 
   minR =find(CumRed>=PixelAmount*lowCut,1,'first');
   minG = find(CumGreen>=PixelAmount*lowCut,1,'first');
   minB =find(CumBlue>=PixelAmount*lowCut,1,'first');
 
   
 
   maxR =find(CumRed>=PixelAmount*(1-highCut),1,'first');
   maxG =find(CumGreen>=PixelAmount*(1-highCut),1,'first');
   maxB = find(CumBlue>=PixelAmount*(1-highCut),1,'first');
 
   
 
   RedMap = linearmap(minR,maxR);
   GreenMap = linearmap(minG,maxG);
   BlueMap = linearmap(minB,maxB);
 
   
 
   imDst = zeros(hei,wid,3,'uint8');
   imDst(:,:,1) = RedMap (I(:,:,1));
   imDst(:,:,2) = GreenMap(I(:,:,2));
   imDst(:,:,3) = BlueMap(I(:,:,3));
 
else
   HistGray = imhist(I(:,:));
   CumGray = cumsum(HistRed);
   minGray =find(CumGray>=PixelAmount*lowCut,1,'first');
   maxGray =find(CumGray>=PixelAmount*(1-highCut),1,'first');
   GrayMap = linearmap(minGray,maxGray);
 
   
   imDst = zeros(hei,wid,'uint8');
   imDst(:,:,1) = GrayMap (I(:,:)); 
 
end
 
%--------------------------------------------------------------------
 
function map = linearmap(low,high)
map = [0:1:255];
for i=0:255
   if(i<low)
       map(i+1) = 0;
   elseif (i>high)
       map(i+1) = 255;
   else
       map(i+1) =uint8((i-low)/(high-low)*255);
   end
end
 
 
 
%-------------------------------------------------------------------
function [I,lowCut,highCut] = parse_inputs(varargin)
 
narginchk(1,3)
 
 
 
I = varargin{1};
validateattributes(I,{'double','logical','uint8','uint16','int16','single'},{},...
             mfilename,'Image',1);
 
if nargin == 1
   lowCut = 0.005;
   highCut = 0.005;
elseif nargin == 3
   lowCut = varargin{2};
   highCut = varargin{3};
else
   error(message('images:im2double:invalidIndexedImage','single, or logical.'));
end

测试驱动代码：

I = imread('bikes1.jpg');
Out = autolevel(I,0.005,0.005); 
imshow([I Out])
 
测试样例：

 

个人感觉：

自动色阶用来做去雾，还是最为稳定的，虽然现在有很多种其他增强方法，例如暗通道去雾，优化对比度去雾，基于Color-Lines的去雾，但是这些算法还不是很稳定，当先验知识失效时，处理失真比较严重。顺便一提的是，美图秀秀的一键去雾即是采用此方法。

--------------------- 
作者：风吹夏天 
来源：CSDN 
原文：https://blog.csdn.net/bluecol/article/details/44732599 
版权声明：本文为博主原创文章，转载请附上博文链接！

https://bbs.csdn.net/topics/20221871

https://blog.csdn.net/xingyanxiao/article/details/48034857

https://www.cnblogs.com/Imageshop/archive/2011/11/13/2247614.html