基于MATLAB的图像⾃适应阈值分割程序最近写了⼀个基于直⽅图的图像阈值分割作业，代码如下：
clear all;
g=imread('Test_Img_1.jpg');
%g=rgb2gray(I);
%thresh(g,99,130,150);
%figure(2)    %绘制直⽅图
[cnts, x] = imhist(g, 256);
[m, n] = size(g);
prob = cnts / m / n;
% sum(prob) == 1
subplot(1,2,1);
plot(x, prob);  title('直⽅图');
y = medfilt1(prob,6);
subplot(1,2,2);
plot(x,y);title('中值滤波');
b=0;
KK=[ ];
T=0;
temp=100;
for k=2:250
Kt=0;
u0=0;
ub=0;
sigma0=0;
sigmab=0;
ind1=find(g<k);
ind2=find(g>=k);
% prob0 =zeros(k,1);
for i=1:k
prob0(i)=size(find(g==i))/size(ind1);
end
for i=1:k
u0=u0+i*prob0(i);
end
for i=1:k
sigma0=sigma0+prob0(i)*(i-u0)^2;
end
% u0=sum(g(ind1)*prob(g(ind1)));
for i=k+1:256
probb(i)=size(find(g==i))/size(ind2);
end
for i=k+1:256
ub=ub+i*probb(i);
end
for i=k+1:256
sigmab=sigmab+probb(i)*(i-ub)^2;
end
gx=1:1:256;
p0=sum(prob(1:k));
pb=1-p0;
pb=1-p0;
fog0=p0*(exp(-(((gx-u0)/sigma0).^2)/2))/sqrt((2*pi))/sigma0; fogb=pb*(exp(-(((gx-ub)/sigmab).^2)/2))/sqrt((2*pi))/sigmab; fog=fog0+fogb;
figure(3)
plot(gx,fog); title('f(g)变化曲线');
pg=p0*fog0+pb*fogb;
pg=pg';
u=unique(g);
u=double(u);
usize=size(u);
matlab直方图
for i=2:usize
if prob(u(i))~=0
Kt=Kt+prob(u(i))*log(prob(u(i))/pg(u(i)));
end
end
KK=[KK Kt];
if(Kt<temp)
temp=Kt;
T=k;
end
end
figure(4)
plot(KK); title('Kt变化曲线');    %绘制Kt变化曲线
axis([0 256 0 4]);
ind3=find(g>=T);
g(ind3)=255;
ind4=find(g<T);
g(ind4)=0;
figure(5)
imshow(g);title('最优阈值分割结果');
disp('>>输出Kt最⼩值');
T

版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。