基于直方图优化的图像增强技术（MATLAB实现）

前言

直方图均衡是一种图像增强的常用手段，但全局直方均衡经常会给图像带来失真等问题，为了处理相关问题，本文采取了全局直方均衡、局部直方均衡和Retinex算法来对图像进行处理。

主函数

img = imread('img.tif');
In1 = RemoveFogByGolbalHisteq(img, 1);
In2 = RemoveFogByLocalHisteq(img, 1);
In3 = RemoveFogByRetinex(img, 1);

全局直方均衡函数

function In = RemoveFogByGolbalHisteq(I, flag)
%I为输入图像
%flag为显示标记
%In为输出图像

if nargin < 2
    flag = 1;
end

%提取R、G、B分量
R = I(:,:,1);
G = I(:,:,2);
B = I(:,:,3);

%分别对R、G、B分量做全局直方均衡
M = histeq(R);
N = histeq(G);
L = histeq(B);

%合成图像
In = cat(3, M, N, L);

%结果显示
if flag
    figure;
    subplot(2, 2, 1); imshow(I); title('原图像');
    subplot(2, 2, 2); imshow(In); title('处理后的图像');
    %灰度画
    Q = rgb2gray(I);
    W = rgb2gray(In);
    subplot(2, 2, 3); imhist(Q, 64); title('原灰度直方图');
    subplot(2, 2, 4); imhist(W, 64); title('处理后的灰度直方图');
end

局部直方均衡函数

function In = RemoveFogByLocalHisteq(I, flag)

if nargin < 2
    flag = 1;
end

%对R、G、B三层向量分别做局部均衡化处理
g1 = GetLocalHisteq(I(:,:,1));
g2 = GetLocalHisteq(I(:,:,2));
g3 = GetLocalHisteq(I(:,:,3));

In = cat(3, g1, g2, g3);

if flag
    figure;
    subplot(2, 2, 1); imshow(I); title('原图像');
    subplot(2, 2, 2); imshow(In); title('处理后的图像');
    %灰度画
    Q = rgb2gray(I);
    W = rgb2gray(In);
    subplot(2, 2, 3); imhist(Q, 64); title('原灰度直方图');
    subplot(2, 2, 4); imhist(W, 64); title('处理后的灰度直方图');
end

function g = GetLocalHisteq(I)
g = adapthisteq(I, 'clipLimit', 0.02, 'Distribution', 'rayleigh');

Retinex算法

function In = RemoveFogByRetinex(I, flag)
%f为输入图像
%flag为显示标记

if nargin < 2
    flag = 1;
end

%提取R、G、B分量
R = I(:, :, 1);
G = I(:, :, 2);
B = I(:, :, 3);

%数据类型归一化
mr = mat2gray(im2double(R));
mg = mat2gray(im2double(G));
mb = mat2gray(im2double(B));

%定义alpha参数
alpha = randi([80, 100], 1) * 20;   %生成一个在80~100之内的数乘以20

%定义模板大小
n = floor(min([size(I, 1) size(I, 2)]) * 0.5);  %模板尺寸为原图矩阵最小边的一半

%计算中心
n1 = floor((n + 1) / 2);
for i = 1:n
    for j = 1:n
        %高斯函数
        b(i, j) = exp(-((i-n1)^2 + (j-n1)^2)/(4*alpha))/(pi*alpha);
    end
end

%卷积滤波
nr1 = imfilter(mr, b, 'conv', 'replicate');
ng1 = imfilter(mg, b, 'conv', 'replicate');
nb1 = imfilter(mb, b, 'conv', 'replicate');
ur1 = log(nr1);
ug1 = log(ng1);
ub1 = log(nb1);
tr1 = log(mr);
tg1 = log(mg);
tb1 = log(mb);
yr1 = (tr1 - ur1) / 3;
yg1 = (tg1 - ug1) / 3;
yb1 = (tb1 - ub1) / 3;

%定义beta参数
beta = randi([80 100], 1) * 1;

%定义模板大小
for i = 1:n
    for j = 1:n
        a(i, j) = exp(-((i-n1)^2 + (j-n1)^2)/(4*beta))/(6*pi*beta);
    end
end

%卷积滤波
nr2 = imfilter(mr, a, 'conv', 'replicate');
ng2 = imfilter(mg, a, 'conv', 'replicate');
nb2 = imfilter(mb, a, 'conv', 'replicate');
ur2 = log(nr2);
ug2 = log(ng2);
ub2 = log(nb2);
tr2 = log(mr);
tg2 = log(mg);
tb2 = log(mb);
yr2 = (tr2 - ur2) / 3;
yg2 = (tg2 - ug2) / 3;
yb2 = (tb2 - ub2) / 3;

%定义eta参数
eta = randi([80 100], 1) * 200;
for i = 1:n
    for j = 1:n
        a(i, j) = exp(-((i-n1)^2 + (j-n1)^2)/(4*beta))/(4*pi*beta);
    end
end

%卷积滤波
nr3 = imfilter(mr, a, 'conv', 'replicate');
ng3 = imfilter(mg, a, 'conv', 'replicate');
nb3 = imfilter(mb, a, 'conv', 'replicate');
ur3 = log(nr3);
ug3 = log(ng3);
ub3 = log(nb3);
tr3 = log(mr);
tg3 = log(mg);
tb3 = log(mb);
yr3 = (tr3 - ur3) / 3;
yg3 = (tg3 - ug3) / 3;
yb3 = (tb3 - ub3) / 3;
dr = yr1 + yr2 + yr3;
dg = yg1 + yg2 + yg3;
db = yb1 + yb2 + yb3;
cr = im2uint8(dr);
cg = im2uint8(dg);
cb = im2uint8(db);

%通过集成处理后的分量得到结果图像
In = cat(3, cr, cg, cb);

%结果显示
if flag
    figure;
    subplot(2, 2, 1); imshow(I); title('原图像');
    subplot(2, 2, 2); imshow(In); title('处理后的图像');
    %灰度画
    Q = rgb2gray(I);
    W = rgb2gray(In);
    subplot(2, 2, 3); imhist(Q, 64); title('原灰度直方图');
    subplot(2, 2, 4); imhist(W, 64); title('处理后的灰度直方图');
end

结果