2D moving sum of a matrix

Hi there

How do I perform a moving sum within a matrix? I.e.: the result of each index will be the sum of all surrounding indexes, as defined by window lengths (with some edge effects, as with a regular moving sum).

I could definitely set up some basic loop function but I anticipate having to do this quite often and so I wish to find a smarter/faster solution. And I suspect that there definitely is a smarter solution. Like maybe using filter2 or something?

Thanks in advance for all help!

Cheers

Jakob

[KSSV]

There is inbuilt function movsum, did you see that?

Answer 1

[Bruno Luong]

Assuming A is your array, w is the window size

% testexample
A=randi(10,5)

movsum(movsum(A,w,1),w,2)

% or
conv2(A, ones(w), 'same')

% or
filter2(ones(w), A, 'same')

[Jakob Sievers]

Thank you so much for this. I knew there was a simple way to do this.

I ended up using fspecial to create a circular sum-filter instead of a basic square sum-filter:

filter=fspecial('disk',5)>0;

Answer 2

[Walter Roberson]

conv2(YourMatrix, ones(WindowHeight, WindowWidth))

You might want to add options such as 'valid' to control the size of output and the edge effects.