Sum particular adjacent elements

I learned about conv2 from the the post:

https://www.mathworks.com/matlabcentral/answers/357208-how-to-sum-all-adjacent-elements-in-a-matrix-with-1-s-and-0-s

My original goal was to sum all the adjacent elements in a matrix. The solution using conv2, from the post above, with the mask:

msk_0 = [1 1 1; 1 0 1; 1 1 1]

worked well.

Using conv2, with a similar logic, now I want to sum the elements above and adjacent to a given element. In this case my mask becomes:

msk_1 = [1 1 1; 0 0 0; 0 0 0] %top

I want to run this on a matrix f, indentical to the mask.

f = [1 1 1; 0 0 0; 0 0 0]

The answer produced is 0

>> conv2(f, msk_1, 'same')

ans =

     0     0     0
     0     0     0
     0     0     0

My expectation was that the mask would result in the sum of the elements above and adjacent to a given element in f. The result in this case would be:

     0     0     0
     2     3     2
     0     0     0

What is it I am missing about the functionality of conv2?

How do I achieve my goal of summing only the specific elements in a given mask?

Thanks for your input.

Answer 1

[Image Analyst]

No, you don't understand what convolution is. With convolution, the kernel gets flipped top-to-bottom, and left-to-right. There are mathematical reasons for this based on linear systems theory. So if you want the sum of the above row, your kernel would have to be

msk_1 = [0 0 0; 0 0 0; 1 1 1] %top

Or you can use imfilter() and not flip the mask because imfilter() does not flip the mask.

[ASC]

Thanks so much!