Image enhancement can be done in the frequency domain by filtering out unwanted frequencies. Computationally, this is done by convolution of a mask and the image of interest.
Before we proceed to enhancing images in the frequency domain, we look into convolution.
Here are images of pairs of dots and squares with corresponding Fourier Transforms.
Here are 2 circles with Gaussian intensity distribution. Increasing variance decreases the frequency-space dimensions of the FT. Also, the patterns are similar to the central regions of the circle and square.
With randomly spaced dirac deltas, the FT appears to be . Convolved with patterns, the resulting FT repeats the patterns in the di
(insert FT of dirac)