清空筛选条件
清空筛选条件

image segmentation problem in satellite image by Graph Cut

1 次查看（过去 30 天）
dakhli mohamed
dakhli mohamed 2019-3-5
Hello
I have a problem with code
when I do the execusion I find a black image but against the resulat estimate is to have an image similar to the image to enter
here is the code and the starting image
  And thank you in advance
function [u, erriter, i, timet] = CMF_Cut
%
% Performing the continuous max-flow algorithm to solve the
% continuous min-cut problem in 2D
%
% Usage: [u, erriter, i, timet] = CMF_Cut;
%
% Inputs: there is no input since all data and parameters can be
% adjusted within the program
%
% Outputs:
% - u: the final results u(x) in [0,1]. As the following paper,
% the global binary result can be available by threshholding u
% by any constant alpha in (0,1):
%
%
%
% - erriter: it returns the error evaluation of each iteration,
% i.e. it shows the convergence rate. One can check the algorithm
% performance.
%
% - i: gives the total number of iterations, when the algorithm converges.
%
% - timet: gives the total computation time.
%
% Example:
% >> [u, erriter, i, timet] = CMF_Cut;
%
% >> us = max(u, beta); % where beta in (0,1)
%
% >> imagesc(us), colormap gray, axis image, axis off;figure(gcf)
%
% >> figure, loglog(erriter,'DisplayName','erriterN');figure(gcf)
%Copyright 2011 Jing Yuan (cn.yuanjing@gmail.com)
s = load('outputMatrix.mat');
outputMatrix = s.outputMatrix;
ur=outputMatrix;
[rows, columns, numberOfColorChannels] = size(ur);
if numberOfColorChannels > 2
% It's a true color RGB image. We need to convert to gray scale.
Igray = rgb2gray(ur);
else
% It's already gray scale. No need to convert.
Igray = ur;
end
%ur = double(imread('cameraman.jpg'))/255;
[rows, cols] = size(ur);
imgSize = rows*cols;
% define the required parameters:
%
% - alpha: the penalty parameter to the total-variation term.
% For the case without incorporating image-edge weights, alpha is given
% by the constant everywhere. For the case with image-edge weights,
% alpha is given by the pixelwise weight function:
%
% For example, alpha(x) = b/(1 + a*| nabla f(x)|) where b and a are positive
% constants and |nabla f(x)| gives the strength of the local gradient.
%
% - cc: gives the step-size of the augmented Lagrangian method.
% The optimal range of cc is [0.3, 3].
%
% - errbound: the error bound for convergence.
%
% - numIter: the maximum iteration number.
%
% - steps: the step-size for the graident-projection step to the
% total-variation function. The optimal range of steps is [0.1,
% 0.17].
%
alpha = 0.5*ones(rows,cols);
cc = 0.3;
errbound = 1e-4;
numIter = 300;
steps = 0.16;
% build up the data terms
ulab(1) = 0.15;
ulab(2) = 0.6;
Cs = abs(ur - ulab(1));
Ct = abs(ur - ulab(2));
% set the initial values
% - the initial value of u is set to be an initial cut, see below.
% - the initial values of two terminal flows ps and pt are set to be the
% specified legal flows.
% - the initial value of the spatial flow fiels p = (pp1, pp2) is set to
% be zero.
u = double((Cs-Ct) >= 0);
ps = min(Cs, Ct);
pt = ps;
pp1 = zeros(rows, cols+1);
pp2 = zeros(rows+1, cols);
divp = pp1(:,2:cols+1)-pp1(:,1:cols)+pp2(2:rows+1,:)-pp2(1:rows,:);
erriter = zeros(numIter,1);
tic
for i = 1:numIter
% update the spatial flow field p = (pp1, pp2):
% the following steps are the gradient descent step with steps as the
% step-size.
pts = divp - (ps - pt + u/cc);
pp1(:,2:cols) = pp1(:,2:cols) + steps*(pts(:,2:cols) - pts(:,1:cols-1));
pp2(2:rows,:) = pp2(2:rows,:) + steps*(pts(2:rows,:) - pts(1:rows-1,:));
% the following steps give the projection to make |p(x)| <= alpha(x)
gk = sqrt((pp1(:,1:cols).^2 + pp1(:,2:cols+1).^2 + pp2(1:rows,:).^2 + pp2(2:rows+1,:).^2)*0.5);
gk = double(gk <= alpha) + double(~(gk <= alpha)).*(gk ./ alpha);
gk = 1 ./ gk;
pp1(:,2:cols) = (0.5*(gk(:,2:cols) + gk(:,1:cols-1))).*pp1(:,2:cols);
pp2(2:rows,:) = (0.5*(gk(2:rows,:) + gk(1:rows-1,:))).*pp2(2:rows,:);
divp = pp1(:,2:cols+1)-pp1(:,1:cols)+pp2(2:rows+1,:)-pp2(1:rows,:);
% updata the source flow ps
pts = divp + pt - u/cc + 1/cc;
ps = min(pts, Cs);
% update the sink flow pt
pts = - divp + ps + u/cc;
pt = min(pts, Ct);
% update the multiplier u
erru = cc*(divp + pt - ps);
u = u - erru;
% evaluate the avarage error
erriter(i) = sum(sum(abs(erru)))/imgSize;
if (erriter(i) < errbound)
break;
end
end
toc
timet = toc
msg = sprintf('number of iterations = %u. \n', i);
disp(msg);
% For example, alpha(x) = b/(1 + a*| nabla f(x)|) where b and a are positive
% constants and |nabla f(x)| gives the strength of the local gradient.
%
% - cc: gives the step-size of the augmented Lagrangian method.
% The optimal range of cc is [0.3, 3].
%
% - errbound: the error bound for convergence.
%
% - numIter: the maximum iteration number.
%
% - steps: the step-size for the graident-projection step to the
% total-variation function. The optimal range of steps is [0.1,
% 0.17].
%
alpha = 0.5*ones(rows,cols);
cc = 0.3;
errbound = 1e-4;
numIter = 300;
steps = 0.16;
% build up the data terms
ulab(1) = 0.15;
ulab(2) = 0.6;
Cs = abs(ur - ulab(1));
Ct = abs(ur - ulab(2));
% set the initial values
% - the initial value of u is set to be an initial cut, see below.
% - the initial values of two terminal flows ps and pt are set to be the
% specified legal flows.
% - the initial value of the spatial flow fiels p = (pp1, pp2) is set to
% be zero.
u = double((Cs-Ct) >= 0);
ps = min(Cs, Ct);
pt = ps;
pp1 = zeros(rows, cols+1);
pp2 = zeros(rows+1, cols);
divp = pp1(:,2:cols+1)-pp1(:,1:cols)+pp2(2:rows+1,:)-pp2(1:rows,:);
erriter = zeros(numIter,1);
tic
for i = 1:numIter
% update the spatial flow field p = (pp1, pp2):
% the following steps are the gradient descent step with steps as the
% step-size.
pts = divp - (ps - pt + u/cc);
pp1(:,2:cols) = pp1(:,2:cols) + steps*(pts(:,2:cols) - pts(:,1:cols-1));
pp2(2:rows,:) = pp2(2:rows,:) + steps*(pts(2:rows,:) - pts(1:rows-1,:));
% the following steps give the projection to make |p(x)| <= alpha(x)
gk = sqrt((pp1(:,1:cols).^2 + pp1(:,2:cols+1).^2 + pp2(1:rows,:).^2 + pp2(2:rows+1,:).^2)*0.5);
gk = double(gk <= alpha) + double(~(gk <= alpha)).*(gk ./ alpha);
gk = 1 ./ gk;
pp1(:,2:cols) = (0.5*(gk(:,2:cols) + gk(:,1:cols-1))).*pp1(:,2:cols);
pp2(2:rows,:) = (0.5*(gk(2:rows,:) + gk(1:rows-1,:))).*pp2(2:rows,:);
divp = pp1(:,2:cols+1)-pp1(:,1:cols)+pp2(2:rows+1,:)-pp2(1:rows,:);
% updata the source flow ps
pts = divp + pt - u/cc + 1/cc;
ps = min(pts, Cs);
% update the sink flow pt
pts = - divp + ps + u/cc;
pt = min(pts, Ct);
% update the multiplier u
erru = cc*(divp + pt - ps);
u = u - erru;
% evaluate the avarage error
erriter(i) = sum(sum(abs(erru)))/imgSize;
if (erriter(i) < errbound)
break;
end
end
toc
timet = toc
msg = sprintf('number of iterations = %u. \n', i);
disp(msg);

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Image Data Workflows 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by