| imconv - 2D convolution |
Outm = imconv(Img, mask [, form]) |
| Img |
| M x N Grayscale (intensity) image in any range. |
| mask |
| n x n matrix, n odd. This is the spatial convolution kernel. |
| Outm |
| The convolved matrix (grayscale image). |
| form |
| determines the size of Outm. It assumes be one of the following values: |
| 'same' |
| Outm has the same size as the original image, M x N. The image is assumed to be 0 outside its bounds. |
| 'full' |
| Outm contains the full convolution, (M+n-1) x (N+n-1). The image is assumed to be 0 outside its bounds. |
| 'valid' |
| Outm contains only the results of the convolution which have been computed within the bounds of Img. Outm will be (M-n+1) x (N-n+1). |
| Function imconv performs 2D spatial convolution of a grayscale image and a mask. |
| edge(Img) Detects edges in Img, using the sobel gradient estimator, 0.5 threshold level and in both horizontal and vertical directions. |
| The other parameters are optional and non-positional. That is, they may be passed to the function by their name. The following example illustrates this. |
initial_dir = PWD;
chdir (SIPDIR + 'images');
// Detect horizontal lines
h = [ -1 -1 -1
2 2 2
-1 -1 -1]
img = imread('gra.jpg');
res = imconv(img,h);
imshow(res,[]);
// Detect diagonal lines
d = [ -1 -1 2
-1 2 -1
2 -1 -1]
res = imconv(img,d);
imshow(res,[]);
chdir(initial_dir);
|
| The kernel is not rotated by 180 degrees. This is in truth a correlation operator, but in practice only symmetric kernels are used. |
| Ricardo Fabbri <rfabbri@if.sc.usp.br>, with help from Scilab Group <Scilab@inria.fr> |
| The latest version of the Scilab Image Processing toolbox can be found at |
| http://siptoolbox.sourceforge.net |
| mkfilter, edge |