Seek help : Convert digital image to frequency domain using fft?

Question

I am doing a thesis "Image Restoration using Inverse Filtering".

I am in the stage to convert digital image from spatial domain to
frequency domain using fft(fast fourier transform). I will program it
using c programming.

I have found fftw as a way to make the fft. However, I am not familiar
with fftw. I compare fft result from Matlab and fftw. They are totally
different.

Is there anyone have skills and experiences in digital image
processing? Thanks.

Answer 1

Hmm, I've used fftw in the past and it looked OK when testing with Matlab (i.e. fftw and Matlab gave the same results). I was only doing a 1-dimensional FFT though.



That's probably not much help, but without knowing more about what you are getting it's tricky to make suggestions (well, not without patronising you... I mean obviously try simple things first - e.g. 1-dimensional real-valued FFT on a sine wave - and compare that between FFTW and Matlab. Ideally load the FFTW answer in Matlab and plot the abs() of the two answers on the same graph)

Answer 2

First, do some experiments with slow Fourier transforms: simply do the numerical integration the hard way. For a laboratory project, that will do fine. (The first scientific program I ever wrote was to do a Fourier transform, back in 1959, before the FFT had even been invented.)



In parallel with your experiments, you can do some research on FFT algorithms. This is not a field that I have followed; the first thing I saw was to do a bit reversal on the table indices, which requires that the number of data points be a power of 2. But the field has exploded; CAT scans rely on the technique.

Answer 3

at the same time as each and all the solutions provide you with a greyscale or B&W image they are going to reason you to loose assessment and factor. In PhotoShop circulate to the image menu and choose "Calculations" decide on purple for the 1st resource and green for the 2d resource. you will now circulate to the mixing pull down, close to the botom of the communique pane, and this could administration the "lighting fixtures" on your image. there is not any set one that works for each image yet play with it until you get one you like or gets you shut. on the backside decide on "new rfile" and hit ok. this could open the recent rfile and it would be an alpha channel. circulate decrease back to image:mode and choose greyscale and once you're printing it on a colour printer turn it to RGB. and make your contrats/brightness variations and whammo you have an exceedingly sparkling sharp B&W image. the image would be crisp and you will no longer get the muddy look desaturatig the image provide you with.