Underexposure vs. High ISO noise

Discussion in 'Nikon' started by narayan, May 5, 2008.

  1. I am using the Nikon 300mm f/4 AF-S lens with TCs on the D200 for photographing
    wildlife. When the available light is low, even the fastest available
    (effective) aperture may not yield a fast enough shutter. Under these
    conditions, is it better to:

    (a) use a low ISO and underexpose the picture and then pull up the EV in post
    processing with Capture NX (upto a max of +2 EV using exposure compensation and
    then levels), or

    (b) use high ISO and expose normally.

    In both cases, there is amplification leading to noise. What are the differences
    in these processes and the character of the noise? Is one approach better than
    the other or are they both equally bad in the sense of removability of the noise
    in post-processing using noise-reduction software like Neat image? Thanks.
  2. I was curious about this myself a while back. I did a quick test with JPEG and photoshop,
    not capture NX so the results might be skewed. I found a lot less noise when boosting the
    ISO in camera rather than a curve adjustment in Photoshop.

    There is no magic bullet for noise removal, i dont like software like neat image. You can
    reduce colored noise by bluring the A&B channels in LAB color space without effecting the
    sharpness of a photo.
  3. FWIW and IMO It's far better to expose as normally as possible. Underexposing at any ISO from what I've seen will lead to more noise than a proper exposure will.
  4. I second Don's observation. I never underexpose on purpose.

    However, I'm a big believer in the empirical method, and in this case the experiment is easy, so I'd suggest trying both ways and see which one you prefer.

    I'd also shoot raw to have the most options for noise reduction.
  5. I would say without doubt or prevarication... it depends.

    Since noise due to underexposure tends to affect shadow areas, it's feasible that the technique you suggest - low ISO, slight underexposure, compensation during post processing - might actually work with certain subjects depending on tonal range. If midtones to upper tones, it might work fairly well. If darker tones, don't count on seeing any advantages.

    High ISO noise tends to be apparent across the frame. Some folks say it affects mostly shadow areas. Mebbe so, but how obvious are those areas? I see noise fairly evenly distributed across all values in test shots online from all dSLRs.

    It's a tough choice. Even my D2H, which is no great shakes by today's standards for high ISO noise, beats the socks off any ISO 800 color film or faster b&w film I've used. The trick is careful, selective application of noise reduction. Go easy on luminance noise reduction, since this affects sharpness. Chroma noise reduction can be applied more liberally, because it mostly affects saturation of certain colors (primarily reds in the D2H).

    If photographing mourning doves at dusk, technique #1 - low ISO, underexposure - might work well with careful post processing.

    Grackles at dusk? I'd go with higher ISO, proper exposure and noise reduction.

    But I don't photograph very sexy or exotic wildlife so YMMV.
  6. There were several discussions about this at DPreview last year. It depends on the specific camera model and processing software - Capture NX or Capture gives the best results with D200. The important improvement is in dynamic range, not noise - upping ISO can push the highest tones over the edge while underexposure allows more headroom at the top. My own tests indicated that noise level of D200 very much depends on actual exposure, not ISO setting. That is also why shadows are much noisier. Setting ISO above 800 (unity gain) for D200 will reduce dynamic range and it is better to underexpose. One of those discussions can be found here.
  7. In theory D200 reaches unity gain at iso800 which means that one electron equals one step in the analog to digital converter. At this point an increase in iso gives basically the same result as an exposure increase in postproduction. For more technical detail check this out: http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/#unity_gain

    So in summary; you should be better off shooting at higher iso up to iso 800 at which point it makes no difference. But it also depends how the exposure increase is done in whatever software you use so as always test yourself and then decide.

  8. Just to clarify, underexposing 1 stop at iso 200 gives a little more noise than shooting at iso 400. Underexposing 1 stop at iso 800 is the same as shooting at iso 1600.
  9. Peter: thanks for the ClarkVision link. I haven't checked his site for new stuff in a while. He seems to do an excellent job measuring and explaining sensor characteristics.

Share This Page