Edge Detection - Customizing Masks

I am trying to apply a Roberts-style edge detection to an image, however I do not want to use the Roberts algorithm supplied by the matlab code i.e. NOT: edge(I,'roberts'). The Roberts masks I need to apply to the image are: x=[1 0;0 -1] y=[0 -1;1 0], which is slightly different that MATLAB's built-in edge(I,'roberts') edge detector.

How can I apply custom Roberts, Sobel, and Prewitt edge detectors in MATLAB?

[PATRICK STANTON]

something like this?:

I = imread('Pic7.bmp');

maskx = [1 0;0 -1];

masky = [0 -1;1 0];

X = imfilter(I,maskx);

Y = imfilter(I,masky);

Z = X+Y;

Answer 1

[Image Analyst]

Yes. That will work. Not sure about the Z though - you might want to take Z = sqrt(single(X).^2 + single(Y).^2) instead. Also, you might want to look into conv2(), which I believe is more highly optimized for speed than imfilter. Be sure to use fliplr and flipud on the masks if the orientation is important though since conv2 flips the kernel.

[PATRICK STANTON]

Thank you for the quick reply! This particular formula only required the sum of the absoulte value of the respective x and y gradients, not a square-root of the sum of squares fortunately. Thank you Obi-Wan.