how to adress neighbours in matrix

Imagine I have matrix A=ones(5). I want to change center of the matrix to 17, A(3,3)=17; In next iteration I want to change all points that are next to this 20 to 5. i will have:

1  1  1  1  1
1  5  5  5  1
1  5 17  5  1
1  5  5  5  1
1  1  1  1  1
In next step i want to change all ones to 10. And so on...

10 10 10 10  10
10  5  5  5  10
10  5 17  5  10
10  5  5  5  10
10 10 10 10  10
Any idea how to do this to large matrix? How to adress only the neighbours but not the points "inside"?

Answer 1

[Image Analyst]

This works for any size matrix, large or not, as long as it's bigger than 5x5. Also you can't get a center element if there is an even number of rows or columns.

A = ones(9)
centerRow = ceil(size(A, 1)/2)
centerCol = ceil(size(A, 2)/2)
A(centerRow-2:centerRow+2, centerCol-2:centerCol+2) = 20;
A(centerRow-1:centerRow+1, centerCol-1:centerCol+1) = 10;
A(centerRow, centerCol) = 17

[Jakub]

But what if i have to do in reverse? I mean first put 17 in center, then tens around, then twenties around...

[Image Analyst]

Why care about what order it's done in as long as it gets done?

[Jakub]

Beacuase i want to use this function in more complicated algorithms dealing with heat spherical transfer. And it has to be computed from the center to the border. This is just a simple example. So I need it in for loop...

[Image Analyst]

You'd have to set the 4 "sides" one at a time. For example to set the outlines at 2 away from the middle:

A(centerRow-2, centerCol-2:centerCol+2) = 20; % Top line
A(centerRow+2, centerCol-2:centerCol+2) = 20; % Bottom line
A(centerRow-2:centerRow+2, centerCol-2) = 20; % Left line
A(centerRow-2:centerRow+2, centerCol+2) = 20; % Right line

If they're not perfect rectangles, but irregular shapes, or if you want the dot to expand out to an octagon or something, then you'd have to use mathematical morphological dilation with a structuring element of the desired shape. See bwmorph() in the Image Processing Toolbox.

Or you could do it by computing the distance transform with bwdist() and then thresholding.

[Jakub]

Thank you very much.