D70, sharp lens and aliasing. Any solution?

Discussion in 'Nikon' started by salvatore.mele, Jul 12, 2005.

  1. After a couple of months with the D70, I think I bumped into aliasing problems.
    This picture has some odd moire-looking colour fringes in thin white structures against a dark background (the boat masts and aerials). I post here a 100% crop of the effect.
    The image was shot as raw, ISO 200, 1/500th, with a 50mm f/1.4 at f/5.6.
    The artefacts are present in the .tif and of course the .jpg both if they are obtained from the .nef through irfanview 3.97 or through Nikon View 6.2.5 as, for instance, discussed here.
    Bob's "pixel peeping" essay makes me think I am observing an aliasing problem, and I've a few questions to the gurus out there:
    • Is this indeed aliasing, or is it a feature of the way the raw information of the sensor is put together into the RGB content of the .tif pixel?
    • If it such a feature, can this be cured with a nef-to-tif conversion other than the one I am using?
    • If it is aliasing, given the MTF in Bob's essay, shall I evict my habit of shooting at 5.6 with my 50/1.4 to try and stay below the Nyquist limit?
    • Your advice and insight are much welcome.
      00Cqk3-24629584.jpg
     
  2. The best raw conversions from D70 files are probably created by Nikon Capture, so you might want to try it out on your photo.
     
  3. This has nothing to do with the sharpness of the lens. I use lenses that are way sharper than the 50/1.4 @5.6 and I do not have any color moire with such lenses.
     
  4. 1) Seems to me that this IS an aliasing issue. But I thought that the aliassing issue and the issue of how a camera converts the r+g+b separate pixel information into the rgb photo you see is closely related to the aliassing issue. <br>
    2) Yes, I think you could use Nikon Capture to do a conversion NEF to TIF that would aleviate this effect.<br>
    3) No, I would still shoot the sharpest image I could, then address any moire or aliassing problems as they crop up using NC. <br>
    Then again, you take sharper photos than I do :).<br>
    It seems to me that it does have to do with a combination of the sharpness of the lens and the very fine line of the mast. The problem occurs when you have high contrast items (bright white stripes on dark, or dark stripes on light background) that approach about one pixel wide (or less) on the CCD. The only solution in camera design would be a "softer" anti-aliassing filter, but then you would lose some of the snappy sharpness. I think that Nikon got it right on the D70 and chose for sharpness, allowing for the occasional moire.<br>
    You solutions are either to use a softer lens (nah, bad idea), recompose so the scene does not have fine lines about the size of a single row of pixel (not always possible), buy a D2X with a finer spacing on the sensor (yeah sure, I would like one too) or to shoot RAW (which you did) and then use Nikon Capture to correct the problem. I'd go for that last solution, and usually shoot raw, except for the snapshots.<br>
    I think you can try Nikon Capture for free for 30 days, so see if it works for you.<br>
    Hope this helps. Seems you know already most of what you need to know.
     
  5. "I use lenses that are way sharper than the 50/1.4 @5.6 and I do not have any color moire with such lenses."


    What Nikon/Nikon compatible lenses do you use that are "way sharper" than a 50mm f/1.4 (AFD?) at f/5.6? Based on Photodo's MTF test, it would be difficult to find a lens that would be noticeably sharper, let alone "way sharper":


    http://www.photodo.com/prod/lens/detail/NiAF50_14D-443.shtml
     
  6. Several, a Printing Nikkor 95mm f/2.8, for example. Once fitted on a focus mount with a matrix chip, EVERYTHING becomes nikon compatible.
     
  7. jbq

    jbq

    -Looks like aliasing to me

    -Not sure

    -When in doubt (and with light and time permitting), it could be a good idea to take an extra shot at f/11.
     
  8. If you want to provoke an aliasing problem you should have a bright thin line of light going at a slight angle across your sensor matrix. - Well thats what you did. Your lens is good enough to expose only one color (sub-)pixel with a high intensity light signal. Good lens :)with great resolution. Also your camera was very steady (and the boat as well).
    This is the price we have to pay unless we tolerate anti-aliasing firm- or software that will introduce "totally different" artefacts.
    Only PS can come to the rescue.
    Cheers
    Walter

    PS: Would be fun to compute the angular resolution. Do you know the diameter of the mast and the wires?
     
  9. Thanks for the feedback. As a follow-up
    • Nikon Capture vs. Nikon View. I had always understood that the algorithms which go from the CCD raw information to the RGB content of each pixel in the .tif where the same in the two programs... Do you (Ikka) mean there are differences exactly in the anti-aliasing department?
    • Angular resolution. The cross section of the mast should be about 25cm and the picture is most likely taken at a distance of 450-500m. This should be about 0.03 degrees...
     
  10. Actually, with a 50mm lens and a horizontal sensor size of 23.7mm, the angle of view is 2xatan(23.7/2x50)=26degrees. With 3008 pixels in the horizontal size, this makes about 0.009 degrees/pixel. The mast, with its 0.03 degrees, should take about three pixels, what it indeed does if I zoom the picture. Now, I find it strange to have hit the limit with something three pixels across...
     
  11. Salvatore, yes, there are differences between the algorithms used by Nikon View and Capture, which mean that you get less moire using Capture. Nikon's position (I asked them) is that View is just for browsing through images, not real work.
     
  12. I can't say of course whether this problem in your case would be solved by using Capture, but it's worth a try.
     
  13. I think I can spot the same problem on the mast of the rightmost boat. Since it is far wider than the original one, I would rule out the aliassing effect. I would take two images of the same scene just to see what are the diferences of the in camera conversion and the conversion made by the two different programs.

    Hope this helps.
     

Share This Page

1111