How do i create a matrix of border angles?

Hello, My input is a matrix of a black-white border image (created with the function border()) i want to create a matrix corresponding to that matrix, that hold the rotation angle values of the border in each pixels (and NaN in the pixels that are not part of the border). for example- if my image is a vertical line, then i get in the line's pixels values of zero, and the rest of the matrix will be NaN. Thanks!

[Image Analyst]

There is no function border() in base MATLAB or the Image Processing Toolbox. Is your border the border of the image - straight lines at the outer edges - or are they boundaries of objects in an image, like you'd get from the edge() function?

[Neta]

sorry, you are right, i meant edge()

[Neta]

any idea how do i create a boundary matrix or how to calculate the gradient with this edge() matrix that i have? thanks a lot!

Answer 1

[Matt J]

Something like this, maybe

   [gx,gy] = gradient(YourImage);

   AngleMap = atan2(gx,gy)*180/pi;  %in degrees, not radians

    AngleMap(~(gx|gy))=nan;

[Image Analyst]

There is a "d" version of atan2 - atan2d - like there is for all the other trig functions. They take or return degrees instead of radians.

[Matt J]

Good to know! But it's new. R2012a and prior releases don't have ATAN2D.

[Image Analyst]

I know people here in answer have asked about it using R2010b, at least sind() link - surprised that they didn't add tangent when they were adding the others.

[Neta]

Hey, thanks for the answer, the problem is that the gradient at the border is zero (it's a local max), and then when i clear the cells around with the 3 code line i left with nothing but zeros. any idea how to make the border pixels itself to contain the angle values?

[Matt J]

Here's another variation which works better,

   gx=diff(YourImage,1,2); gx(:,end+1)=0;
   gy=diff(YourImage,1,1); gy( end+1,:)=0;

     AngleMap = atan2d(gx,gy)+90; 
      AngleMap(~YourImage)=nan

Answer 2

[Image Analyst]

Why not just use imgradient() in the Image Processing Toolbox. It gives you both gradient magnitude and gradient direction, like this example from the help shows:

% Read image and compute gradient magnitude
% and gradient direction using Prewitt's gradient operator.
  I = imread('coins.png');
  [Gmag, Gdir] = imgradient(I,'prewitt');
  figure
  imshowpair(Gmag, Gdir, 'montage');
  title('Gradient Magnitude, Gmag (left), and Gradient Direction, Gdir (right), using Prewitt method')
  axis off;

[Neta]

i don't have the image processing toolbox.. any idea how to implement such a thing?

[Image Analyst]

You can use conv2() with kernels oriented in the vertical and horizontal direction: [-1,1] and [-1;1].