前言

MATLAB进行图像处理相关的学习是非常友好的，可以从零开始，对基础的图像处理都已经有了封装好的许多可直接调用的函数，这个系列文章的话主要就是介绍一些大家在MATLAB中常用一些概念函数进行例程演示！

OSTU又叫大津法, 由日本学者大津于1979年提出。

（1）从大津法的原理上来讲，该方法又称作最大类间方差法，因为按照大津法求得的阈值进行图像二值化分割后，前景与背景图像的类间方差最大。

（2）它是按图像的灰度特性，将图像分成背景和前景两部分。因方差是灰度分布均匀性的一种度量, 背景和前景之间的类间方差越大,说明构成图像的两部分的差别越大, 当部分前景错分为背景或部分背景错分为前景都会导致两部分差别变小。因此, 使类间方差最大的分割意味着错分概率最小。

一. MATLAB仿真

clc;clear;close;

tic
[filename,pathname]=uigetfile('*.jpg;*.png;*.bmp','select the file'); 
im=[pathname,filename]; 
A = imread(im);
imagBW = otsu(A); 
figure(1),imshow(A),title('原图');
figure(2),imshow(imagBW),title('otsu阈值二值化分割图');
toc

function imagBW = otsu(imag) 
imag = imag(:, :, 1); 
 
[counts, x] = imhist(imag);  % counts are the histogram. x is the intensity level. 
GradeI = length(x);   % the resolusion of the intensity. i.e. 256 for uint8. 
varB = zeros(GradeI, 1);  % Between-class Variance of binarized image. 
 
prob = counts ./ sum(counts);  % Probability distribution 
meanT = 0;  % Total mean level of the picture 
for i = 0 : (GradeI-1) 
    meanT = meanT + i * prob(i+1); 
end 
varT = ((x-meanT).^2)' * prob;  
% Initialization 
w0 = prob(1);   % Probability of the first class 
miuK = 0;   % First-order cumulative moments of the histogram up to the kth level. 
varB(1) = 0; 
% Between-class variance calculation 
for i = 1 : (GradeI-1) 
    w0 = w0 + prob(i+1); 
    miuK = miuK + i * prob(i+1); 
    if (w0 == 0) || (w0 == 1) 
        varB(i+1) = 0; 
    else 
        varB(i+1) = (meanT * w0 - miuK) .^ 2 / (w0 * (1-w0)); 
    end 
end 
 
maxvar = max(varB); 
em = maxvar / varT  % Effective measure 
index = find(varB == maxvar); 
index = mean(index); 
th = (index-1)/(GradeI-1) 
imagBW = im2bw(imag, th); 
 
% thOTSU = graythresh(imag) 
% imagBWO = im2bw(imag, thOTSU); 
end