for loop, assign variable

Hi, I am trying to assign distance between each pixel and picked point on mri to variable ed in for loop. However all elements are same. I dont understand the reason. Thank you...

clc;
clear all;
a=imread('tumor.jpeg');
b=imresize(a,3);
c=im2bw(b);
imshow(c);
d=zeros(size(c));
%150,210

[ptx,pty] = ginput(1);
newdata=[ptx, pty];
n=25;
%[count,x] = imhist(c);
[rows, columns]=size(c);
  
 for i=1:rows
     for j=1:columns
   
     %calc and store sorted euclidean distances with corresponding indices
       ed(i)=sqrt(sum((newdata-c(i,j)).^2));
       euc=ed';
     end
end

[Bjorn Gustavsson]

Your newdata are a selected point in the image, while c is the intensity (grayscale) of your pixels. This looks like a bug to me. I guess that you either want the length between th image pixels and your selected pixel, that would be norm(newdata-[j,i]), or the grayscale difference between the pixels and the pixel closest to newdata, that would be c(round(newdata(2)),round(newdata(1)))-c(i,j).

Answer 1

[KSSV]

You should initiliaze ed before the loop.

 
 for i=1:rows
     for j=1:columns
   
     %calc and store sorted euclidean distances with corresponding indices
       ed(i,j)=sqrt(sum((newdata-c(i,j)).^2));       
     end
end

Loop is not required here.

ed=sqrt((newdata(1)-c).^2+(newdata(2)-c).^2);

[KSSV]

Ahsen Feyza Dogan commented:

Thank you but it does not work true. I want to calculate distance between picked point and all pixels. Image size is 528*528 so distance array should contain 528 element. In my code all of them is same but it shoud not be.

[KSSV]

YOur c is a matrix......a point subtracted from matrix..will give you a matrix. How it will be of only size 528?

[KSSV]

Try:

a=imread('image.jpeg');
b=imresize(a,3);
c=im2bw(b);
imshow(c);
d=zeros(size(c));
%150,210
[ptx,pty] = ginput(1);
newdata=[ptx, pty];
n=25;
%[count,x] = imhist(c);
[nx,ny] = size(c) ; 
[X,Y] = meshgrid(1:nx,1:ny) ; 
ed=sqrt((newdata(1)-X).^2+(newdata(2)-Y).^2);

Answer 2

[Alex Mcaulley]

As Bjorn Gustavsson said your code make no sense. To calculate distances you have to use the coordinates of each pixel, not mixing coordinates and intensity values. One option without loop:

a = imread('tumor.jpeg');
b = imresize(a,3);
c = im2bw(b);
imshow(c);
d = zeros(size(c));
%150,210
[ptx,pty] = ginput(1);
newdata = [ptx, pty];
n = 25;
%[count,x] = imhist(c);
[rows, columns] = size(c);

 crows = repmat((1:rows)',1,columns);
 ccolumns = repmat(1:columns,rows,1);
 euc = sqrt((newdata(1) - crows).^2 + (newdata(2) - ccolumns).^2)

[KSSV]

crows = repmat((1:rows)',1,columns);
 ccolumns = repmat(1:columns,rows,1);

The above is not good. Use meshgrid

[Alex Mcaulley]

Why is not good? The code works and the execution time is similar than using meshgrid (In fact, my code results in a slightly better performance in my computer).

%5 different executions
nx = 512;
ny = 512;
timeit(@() meshgrid(1:nx,1:ny))
timeit(@() repmat((1:nx)',1,ny)) + timeit(@() repmat(1:ny,nx,1))
%1st 
ans =

    0.0013
ans =

    0.0011
%2nd
ans =

    0.0015
ans =

    0.0011
%3rd
ans =

    0.0014
ans =

    0.0012
%4th
ans =

    0.0014
ans =

    0.0012
%5th
ans =

    0.0013
ans =

    0.0013

Answer 3

[Ahsen Feyza Dogan]

Since it is distance, there is no x and y coordinate. There shoud be 278784 elemnt from 528*528

[Bjorn Gustavsson]

Yeah, about that. You have one selected point, lets call that one r_slected and an image, lets call that one Img. Now, from each image pixel you want the distance to r_selected. You get that with KSSV's second suggestion, or Alex's, lets repeat it here:

[ny,nx] = size(Img); 
[X,Y] = meshgrid(1:nx,1:ny); 
l2slected_point = sqrt((r_selected(1)-X).^2+(r_selected(2)-Y).^2);

That will make l2selected_point a double array of size 528 x 528 (or whatever size your image has), which is your 278784 distances - but you will have them in an order that is sensibly arranged to be of the same size as your image, which will be crucial for further use.

Try this code and before doing anything more display the result:

imagesc(l2selected_point),colorbar