A 2D convolution can be made in two step in most of the case (by using separable filter). Look to wikipedia page for sobel edge detector (sobel_operator) for example.
So you do it with the 1D convolution code provided as exemple by changing the gaussian kernel into a sobel kernel [1 2 1]T * [+1 0 -1]
So you can apply horizontal kernel and transpose the image if you need. Or write a method to...
2009-07-06 13:41:29 UTC in Adobe Source Libraries
