home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.image.processing
- Path: sparky!uunet!europa.asd.contel.com!gatech!swrinde!cs.utexas.edu!asuvax!ncar!csn!news.den.mmc.com!iplmail!dans
- From: dans@iplmail.orl.mmc.com (Dan Sullivan)
- Subject: Re: need help with FFT's...
- Message-ID: <1993Jan21.175604.9744@iplmail.orl.mmc.com>
- Sender: dans@iplmail (Dan Sullivan)
- Organization: Martin Marietta
- References: <1993Jan20.184425.12744@ms.uky.edu> <l8h39wm@rpi.edu>
- Date: Thu, 21 Jan 1993 17:56:04 GMT
- Lines: 47
-
- In article <l8h39wm@rpi.edu>, raffuj@rebecca.its.rpi.edu (Jose Raffucci) writes:
- |>
- |> I'm trying to add a low pass filter so I can check too see if it's
- |> doing what it's supposed to. It's not. I'm getting what appear to be
- |> the upper edges of some sinc waves, especially in the darker portions
- |> of the pictures. I've checked this out with another program I found
- |> that fft's images, and the same thing happens.
-
- That sounds exactly like it should. Since you are doing an *ideal*
- low-pass filter (all 1's in the pass band, all zero's in the stop band),
- you will indeed get a diffraction pattern in your output image. You are
- effectively convolving the image with a sinc function, which causes the
- ringing.
-
- If you are trying to simulate lower-resolution optics, this is correct.
-
- You might try a smoother filter function that falls off more slowly
- at higher frequencies. A gaussian is one idea.
- |>
- |> Related question:
- |>
- |> What should I do with the data that dips below zero when I
- |> return from the frequency domain? I Need to keep the intensity values
- |> of the pixels, since they're linearly mapped to the signs' luminance.
- |> That pretty much eliminates moving everyting up, and making the lowest
- |> point the new zero. any suggestions?
- |>
- Um, I'm not an optics type, but wouldn't luminance be the magnitude^2
- of the output image? That would take care of any negative pixels.
-
- |> Now, before anyone suggests I not do the fft and move over to a
- |> sliding kernel method, let me mention that if I could I would. The
- |> problem is that I don't know how to change the filter functions to
- |> their corresponding kernels.
- |>
- The FFT is just a convenient computational technique; if you do it right
- you'll get the same results as the space-domain implementation. It sounds like
- you are doing it right.
-
- --
- Dan o
- ____ o o
- Dan-Sullivan@orl.mmc.com (_/\_)o
- Martin Marietta Electronic Systems ()=o
- MP 1160, Orlando, FL USA
- "You keep using that word. I do not think it means
- what you think it means."
-