How to extract and/or average diagonal pixels in an image?

Looking for some advice here. I have a 2D image which contains a slightly tilted (theta ~ 3 degrees) sinusoidal pattern (e.g. monochromatic interference fringes), and would like to take the mean across each diagonal of the fringe pattern, for a large range in one of the image dimensions. Is there a way to extract each diagonal? And maybe scan the diagonal extraction across the image? I've attached a simple concept and code to generate the pattern. Thanks in advance.

% Tilted Sinusoidal Pattern

N = 256;

x = 1:N; y = x;
[x0,y0] = meshgrid(x,y); % Set image grid

theta = pi/2 - 3*pi/180; % angle of tilt in pattern

I = 1 + cos(2*pi*(x0*cos(theta)+y0*sin(theta))); % Generate interference pattern

imagesc(I),
colormap(gray),
title('Simulated Fringe Pattern')

[Walter Roberson]

How do you define "diagonal" for this purpose?

Answer 1

[John BG]

1.- Acquiring image

A=imread('fringes1.jpg')
figure(1);imshow(A)

what you call a diagonal is a cross section. You want a horizontal cross section, but it can be defined for any 2 points:

p=ginput(2)

check

   figure(1);hold all; plot(p(:,1),p(:,2),'ro')

2.- find the amount of pixels between points p(1,:) and p(2,:) in straight line

N=max(abs(p(1,1)-p(2,1)),abs(p(2,1)-p(2,2)))

3.- cross section points:

Lx=linspace(p(1,1),p(2,1),N)
  Ly=linspace(p(1,2),p(2,2),N)

check

figure(1);hold all; plot(Lx,Ly,'r')

4.- extract the image values across the cross section line:

indexing images requires inputs of type uint

Lx2=uint64(floor(Lx))
Ly2=uint64(floor(Ly))

CS1_R=A(sub2ind(size(A(:,:,1)),Lx2,Ly2))
CS1_G=A(sub2ind(size(A(:,:,2)),Lx2,Ly2))
CS1_B=A(sub2ind(size(A(:,:,3)),Lx2,Ly2))
nCS1=[1:length(CS1_R)]

and the result are 3 cross sections, one each primary color

figure(3);subplot(3,1,1);plot(nCS1,CS1_R,'r')
          subplot(3,1,2);plot(nCS1,CS1_G,'g')
          subplot(3,1,3);plot(nCS1,CS1_B,'b')

To get horizontal cross sections set y constant.

If you find this answer of any help solving your question, please click on the thumbs-up vote link,

thanks in advance

John

[e-mail]