imfilter

Hi guys,

I'm confused by the optimal parameters, 'corr' and 'conv'.

Assuming there is a matrix A = [1 2; 8 9; 0 8], a filter f = [-1 -1; 2 2; -1 -1];

It seems that 'corr' and 'conv' will make same result.

I kind of understand 'conv'. It's convolution for the concrete case. But for 'corr', I'm confused. Does it mean the function 'corr2'? I have checked. it seems no.

Can you guys give an example to explain it?

Thanks,

Zhong

Answer 1

[Wayne King]

Zhang, the difference is that there is a rotation of the kernel by 180 degrees (pi radians) between the two.

So for:

   h = [1     2     3
       4     5     6
       7     8     9];

and

   h1 = [9 8 7
           6 5 4
           3 2 1];

 x = randn(20,20);
 isequal(imfilter(x,h,'corr'),imfilter(x,h1,'conv'))

returns a 1.