Linear filtering: Convolution

Convolution
is defined as g(x,y) = h(x,y) * f(x,y)
or G(u,v)=H(u,v) x F(u,v) in the Fourier domain
Here you can create the kernel of the filter h. The number of elements
is 3x3 or 5x5 depending on your choice.
The kernel can be created different ways:
· "Fill with 0/1" to fill the coefficients with 0s or 1s
· Enter your own coefficients
In the later case, several options are also available:
· "Free coefficients": you can edit all of 9 (or 25) coefficients
· "Symmetry": only the first quarter (in yellow and green) of the kernel
can be edited. The rest of the kernel is filled automatically so as to
create symmetry on both horizontal and vertical axes
· "Horizontal anti-symmetry": in that case, coefficients in the 'right'
part of the kernel are opposite sign of those in the left
· "Vertical anti-symmetry": same thing but with top/bottom parts of the
kernel

|