Nikon D80 Digital Camera Night Image Noise Tests

Discussion in 'Nikon' started by Wayne J. Cosshall, Aug 11, 2007.

  1. Wayne J. Cosshall, Aug 11, 2007
    1. Advertisements

  2. Wayne,

    Thanks for publishing those. I am confused by your results.

    Taking the ISO 1600, for example, to me the NR OFF image (upper) has less
    noise, but is marginally less sharp then the NR ON image (lower).

    Could you please explain in a little more detail what makes you say: "At
    1600ISO noise reduction does help...". Where do you see this in the lower

    Do you have an equivalent D40 test?

    David J Taylor, Aug 11, 2007
    1. Advertisements

  3. HI David,

    I write my remarks looking at the RAW files, so the JPEGs may not fully
    reflect what I see onscreen. That's why the RAW files are there for
    download if people want them. It is a tradeoff of wanting the page to
    load reasonably fast and having illustrations. What I found (which is
    typical of most cameras) is that the noise reduction feature really does
    not help that much and sometimes makes it worse. Though at 1600 on the
    D80 it did seem to a bit. Generally all it does really is blur the detail.

    Now, I'll go recheck those just in case I got them swapped around when I
    was checking them, but I don't think so.

    D40x ones are coming. Tests done, just have to write them up and upload.
    Expect in the next couple of days (life gets in the way sometimes :).



    Wayne J. Cosshall
    Publisher, The Digital ImageMaker,
    Photography and Art Forums
    Personal art site
    Wayne J. Cosshall, Aug 11, 2007
  4. Wayne J. Cosshall, Aug 11, 2007
  5. Wayne J. Cosshall

    ASAAR Guest

    Even with my slow dialup connection, I would have preferred if you
    had not stripped the EXIF data. It would have made the jpg crops
    much more useful for evaluating things not mentioned in the text.
    ASAAR, Aug 11, 2007
  6. Thanks, Wayne. Noted that you were looking at raw rather than JPEG. It
    was the verbal description of what changes /you/ were seeing I was after -
    detail on low-contrast areas, edge detail, noise level, fine-grained
    versus coarser grained noise etc.

    By the way - my take is the even at 1600 ISO the pictures are perfectly
    usable, but I don't print large and a bit of noise never bothered me,
    unlike others.

    David J Taylor, Aug 11, 2007
  7. Wayne J. Cosshall

    Pete D Guest

    My thoughts exactly for most shooting.
    Pete D, Aug 11, 2007
  8. Wayne J. Cosshall, Aug 11, 2007
  9. Wayne J. Cosshall

    acl Guest

    If the NR you refer to is "Long exposure NR" (ie dark frame
    subtraction, where the camera takes a second, dark exposure and
    subtracts it), it's not supposed to remove the noise that you see
    there (if anything, it increases its variance by sqrt(2)). Its purpose
    is to remove fixed noise, such as cause by heat sources near the
    sensor (switch off this NR and take a 5min exposure at ISO 1600 and
    you'll definitely see what I mean). I can find online samples if you
    acl, Aug 11, 2007
  10. Wayne J. Cosshall

    Paul Furman Guest

    I checked out your raw files for the ISO 100 shot. I converted with
    increased contrast & brightness & EC+2 then because of differing color
    casts, I desaturated & did screen shots zoomed to 200% so it's easier to
    see the pixels:
    I see better detail in the NR on version.
    Paul Furman, Aug 11, 2007
  11. Here are some 5 minute ISO 800 exposures comparing a Canon
    and Nikon:

    I agree that the NR step increases noise in the darkest
    parts of the image by root 2. For brighter parts of the image,
    the dominant noise is from photon noise (Poisson statistics),
    so should appear about the same in NR on as NR off.

    Roger N. Clark (change username to rnclark), Aug 11, 2007
  12. Wayne J. Cosshall

    Paul Furman Guest

    I think it increases noise and detail in the shadows.

    In your tests, Figure 3.A at the bottom shows the Nikon noise much much
    sharper which reinforces what I've heard that a lot of the Canon noise
    advantage is the stronger anti-aliasing filter or in this case I guess
    it's something in the post-processing with a similar effect. Also, note
    that this is comparing a $7,000 Canon with a $500 Nikon.
    Paul Furman, Aug 11, 2007
  13. Wayne J. Cosshall

    acl Guest

    These are both without dark frame subtraction, right?
    acl, Aug 11, 2007
  14. Wayne J. Cosshall

    acl Guest

    Actually is there anywhere I can get a raw file of a dark frame for eg
    a 1D (or whatever canon)? I'm curious to look at its fourier transform
    and compare to my camera's (it's a slow day here...).
    acl, Aug 11, 2007
  15. Wayne J. Cosshall

    ASAAR Guest

    Based on many of your previous replies I was hoping for a more
    complete response. I guess we'll see how future images are produced.
    ASAAR, Aug 11, 2007

  16. Here are the raw images from the above web page:
    (6.3 megs)
    (5.5 megs)

    It's now a slow day here too, after staining my deck in the
    hot sun.

    Roger N. Clark (change username to rnclark), Aug 11, 2007
  17. That's what I said.
    The raw noise is single pixel so should not look anything but
    single pixels. I think what you are seeing is the differences
    in the raw converters (nikon's engine versus canon's)
    in the raw to tif conversion.

    The 1D Mark II was a $4,500 camera, not 7,000, and is relatively
    old technology relative to Canon's newer sensors. For example,
    the bottom of the canon DSLR line, the XTi, reportedly
    (in the amateur astronomy forums) has even better darks,
    with lower fixed pattern noise, and less amplifier glow,
    same with the new 1D Mark II.

    I have other cameras to post, like 30D and they are similarly
    very good. You can also see differences in the read noise
    between cameras (mostly CCDs have higher read noise than
    canon's CMOS) at:
    see Figure 3.
    I'll add more dark frames when I get time.

    Roger N. Clark (change username to rnclark), Aug 11, 2007
  18. Yes.
    Roger N. Clark (change username to rnclark), Aug 11, 2007
  19. Wayne J. Cosshall

    acl Guest

    Great thanks! If my program doesn't choke at the mass of data (it's in
    Lisp, and I never intended it to work with images when I wrote it...)
    I'll post the results for anybody interested.
    acl, Aug 11, 2007
  20. Wayne J. Cosshall

    Paul Furman Guest

    OK thanks, I converted your raw files in CS1 ACR with the same settings
    and got this from a center crop:
    Paul Furman, Aug 11, 2007
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.