View Single Post
  #11  
Old 04-10-2016, 11:48 PM
Maurice's Avatar
Maurice
Registered User

Maurice is offline
 
Join Date: Jun 2006
Location: Melbourne
Posts: 436
If the ripple frequency is constant, but the effect is displaced from frame to frame, then fft filtering can work.
Attached is a mars image from a few years ago.

'Original mars' is the frame as stacked & processed.

'Noisy mars' is the same frame with a fixed pattern noise applied to simulate your problem

'fft noisy mars' is the frequency map (real component) of the noisy image. You can clearly see the frequencies where the pattern resides (hot spots).

'masked fft frequencies' is the same map with the offending frequencies masked.

'fft filtered' is the inverse fft of the frequency map showing the reconstructed image with the masked frequencies. Its not perfect, but it is a big improvement on the 'noisy' image. With a little care in the masking applied, a better result can easily be obtained.

It is fairly close to the original, pre-noise image.
All processing done in IRIS.

cheers
Maurice
Attached Thumbnails
Click for full-size image (original mars.jpg)
37.1 KB28 views
Click for full-size image (noisy mars.jpg)
42.0 KB24 views
Click for full-size image (fft noisy mars.jpg)
59.3 KB20 views
Click for full-size image (masked fft frequencies.jpg)
36.8 KB20 views
Click for full-size image (fft filtered.jpg)
50.9 KB25 views
Reply With Quote