A Question About Noise and Prints

A reader asks:

I’m curious if you know the answer.  I don’t print very often so I’m not experienced with it.  I keep reading about noise in digital of course.  I also have read many remarks about noise not being visible in print under some conditions.  So I’m curious if there is a threshold.  For example, on the 5D2, what is the print size where noise becomes visible at a given ISO?  The 7D?  I realize that where it becomes objectionable may be subjective, but I thought it may be interesting to know when it is visible or invisible in print.

Let’s assume other technical details are handled well… good exposure, good focus, sharp lens, and no camera shake.

Thanks for writing. This is a great question and a subject that lots of people worry about. I’m not sure I know the answer, but I have an answer based on my own experience. The situation turns out to be a bit fuzzy in the end – in other words, if there is a threshold a number of factors could change your notions of where it might be. (I welcome comments from others who have experience to share.)

From reading photography discussions one could get the idea that image noise in digital photography is a terrible and limiting problem. Discussions often focus on questions like “which camera produces less noise?” and “how do I fix this noise problem.” We see 100% magnification crops of images in which noise is, indeed, quite visible. There are most certainly noise issues that we have to concern ourselves with, but all too often people get worked up over noise issues that are insignificant or even imperceptible – and which are often easy to resolve.

First, every DSLR image contains noise. There is really no such thing as a truly “noise free” DSLR photograph. The object is not to do the impossible and eliminate noise but rather to control it to an extent that it is not a problem in the final image. Since most photographs do not become large prints but are instead shared as relatively small jpg files on the web or in email or are printed at letter-size or smaller dimensions, even large amounts of noise may not be visible at all.

But that wasn’t your question. You are more interested in what happens as print size increases. My answer to this has several parts and it, as mentioned earlier, based more on my personal experience than anything else:

  • A good place to start the consideration of “acceptable noise” is perhaps by looking at prints made from film. I think people who first became serious about photography in the digital era are unaware of the “noise” levels in some of the great photography made with the older technology. If you look into this – a visit to a good museum with classic photographic prints is a good start – you’ll find that a) it was extremely rare to make prints from 35mm film that are anywhere close to the sizes we produce today from DSLRs, b) most prints at these large sizes were probably made from large format film originals, and c) you can clearly see the film grain in these prints. Unless we are ready to dismiss the greatest work by past photographers using film because it is too “noisy” (from grain) then we might consider reevaluating the notions that “noise free” is necessary or realistic.
  • Other issues will likely limit print size before noise becomes the limiting factor. With current full frame DSLRs, excellent lenses, and great care in shooting and post processing it is possible to make great prints at 24″ x 36″ sizes that don’t fail in any of the ways that prints fail as they become larger. With care, even larger prints are possible. But at some point the image quality will not hold up, at least not to a “fine art” print standard. But in most cases it is some other element besides noise that defines this point of “too big” – perhaps fine details start to be “smeared” or the focus is not quite as sharp as you would like or you feel that you have pushed the up-rezzing process too far. To be direct, I think print size will likely be limited by other things before it is limited by noise.
  • In any case, a little bit of noise is not always a bad thing. In fact, in some cases it can be a good thing! Some examples are obvious – many people like the “edgy” retro quality of the grain from high speed black and white films. In other cases a bit of noise can create a richer texture than you find in an image where the photographer has used noise reduction techniques in post.
  • Not all noise is created equal. Noise can come from a number of sources including: shooting at high ISO, underexposing, dark areas in very wide dynamic range images, trying to excessively push dark areas in post, very long exposures, and so forth. If you know you are going to make very large prints, there are decisions you can make when you shoot and when you post-process that can anticipate and reduce these issues. Some examples:
    • Follow the “shoot to the right” mantra in most cases. Pushing luminosity data into the “bright” part of the histogram can produce smoother gradients and less noise. (But avoid blowing out the brightest tones.)
    • In situations with very high dynamic range several techniques can handle this without losing shadow detail and picking up additional noise there. These include using graduated neutral density filters and exposure bracketing.
    • Use the long exposure noise reduction feature for long exposures such as night photography. It does increase the time required for each shot, but it makes a significant difference on long exposures.
    • Avoid having to push shadows (via extreme fill, black point, shadow/highlight, exposure, or curves adjustments) in post. By design, DSLRs allocate fewer data points to the dark end of the brightness spectrum. When you try to push very dark portions of the image back into the middle range to recover detail you run the risk of getting noise and other artifacts. (A hint – if this happens you can sometimes mask it by selectively adding a bit of noise to these areas!)
  • The subject makes a difference. Noise is far less visible in a photograph with high levels of fine detail which tend to mask it. On the other hand noise is more likely to be noticed in images containing large areas of uniform or near uniform color and brightness. (Fortunately, such areas usually respond better to noise reduction techniques than do the “fine detail” areas.
  • Sometimes compromises are necessary. What happens if you just must shoot at ISO 6400 and you have a wide dynamic range scene? You get noise! And lots of it! Here your options probably include “embracing the noise” and making it part of your concept, trying to reduce it in post – even at the expense of losing some detail, using a lower ISO and risking underexposure or subject motion. Until we have cameras with infinitely high ISO ratings and infinitely low noise levels (ain’t gonna’ happen!) we’ll have to “just deal” with some situations.

Since you asked about the 5D2, here is my current thinking based on shooting at various ISOs. At ISO 100, I think that the print size limitations of a well exposed shot will come from other factors before they come from noise. I feel that ISO 200 is generally more or less indistinguishable from ISO 100 in terms of noise. I see a bit more noise at ISO 400, but to me it seems pretty minor and wouldn’t likely limit print size – but if it did a bit of NR in post should deal with it. At ISO 800 it is more noticeable – I can see it at 12 x 18 if I look carefully – but the quality of the noise is often good and I’d just go with it. Depending upon the subject I think that you’ll see the effects of noise more clearly at ISO 1600 – it is hard for me to image shooting a landscape or architecture at that ISO with this in mind, but I imagine that the right sports image would be fine. I really haven’t shot much at higher ISO as of this writing.

(Notes: I usually try to shoot at ISO 100 – if you know my subjects you’ll understand why. However, I regularly use ISO 200 and I don’t shy away from ISO 400 when I need it. I rarely used 800 or 1600 but do at times. I print on the Epson 7900 which has a 24″ carriage width.)

3 thoughts on “A Question About Noise and Prints”

  1. Charles, thanks for the point about paper – it certainly is true that it will be less visible, and perhaps less objectionable if you do see it on matte papers. Like you, I think that the chroma (“color”) noise is more problematic than luminance (contrast or brightness) noise. The latter may suggest something more like film grain in some cases.

    Dave, I don’t have any experience with the 7D, so thanks for sharing yours.

    Dan

  2. Great article Dan! I have the 7D and find ISO 800 is really pushing it for noise. I have also found that Lightroom and ACR doesn’t handle the RAW conversion very well for the 7D. Converting the RAW file with Canon’s DPP or Capture one gives a bit better results, but the noise levels in the 5D is still better (as expected).

  3. I think a couple of other factors need to be pointed out. One, it also depends on what type of paper you are using to make prints. Noise is most prominent on glossy paper. At least to me, it is. Also, many labs reduce the ppi when a print reaches somewhere around 11X17 in size. They do this, because people will not be “Pixel peeping” and will be viewing the image from farther away. Therefore, it is not as critical.

    Great write-up though! I just changed my noise reduction method and I am no longer reducing the contrast noise. I am only reducing the chroma (color) noise. The chroma noise is what always bugs me. One last notion is it’s not so much the presence of noise, but the pattern that it leaves. The new Mark IV has a film-like pattern. So again, for me, I can live with noise, if I can eliminate chroma noise and it doesn’t have banding.

Join the discussion — leave a comment or question. (Comments are moderated and may not appear immediately.)

This site uses Akismet to reduce spam. Learn how your comment data is processed.