How Important Noise Reduction for Black-and-White?

[JimAscher]

I am somewhat puzzled for the significant concern voiced over a camera's inherent ability to shoot at higher ISO's with less noise, or the need to employ special noise reduction software, when the intended outcome is to produce a black-and-white print.  Isn't the net result of producing a black-and-white print with a higher level of RAW noise that merely(!) of having a print with an increased "grainy" effect, similar to various black-and-white films?  I always hav e the feeling in posing questions in this wonderful forum that I' asking really too basic questions, although no respondents to my queries have ever (yet) made me feel woefully (and inexcusably) ignorant.

[john beardsworth]

Just because the image is output as black and white doesn't mean it has to ape film graininess! Have you never heard of APX25 or PanF? Equally, high ISO colour film could produce interesting grainy effects.

[Mike Arst]

I am somewhat puzzled for the significant concern voiced over a camera's inherent ability to shoot at higher ISO's with less noise, or the need to employ special noise reduction software, when the intended outcome is to produce a black-and-white print.  Isn't the net result of producing a black-and-white print with a higher level of RAW noise that merely(!) of having a print with an increased "grainy" effect, similar to various black-and-white films?  I always hav e the feeling in posing questions in this wonderful forum that I' asking really too basic questions, although no respondents to my queries have ever (yet) made me feel woefully (and inexcusably) ignorant.

When I first used a 1.3-megapixel camera (there really were such things, honest :-) I thought I'd need to be concerned about noise. I never was. It just wasn't an issue. All the noise-related torture people would go through in sundry photo forums (though not typically this one) struck me as very odd. The kind of noise you might see during some marathon pixel-peeping session with nose pressed to screen just never seemed to materialize in some significant way in a print.

Sometimes people e-mail my wife and me a photo shot with some low-life :) cigarette-pack-sized camera. As I view the image on the screen I think Eeuuwww... look at all that noise. Gross. But if I make a small print out of it just for taping to the door of the fridge -- where's the noise? Nowhere to be seen (except perhaps under a magnifying-glass, but so far we've avoided the temptation to look at images that way).

If I were to print black and white images taken with the high-ISO-capable digital SLR I would probably be somewhat disappointed precisely because they surely won't have any grain to speak of. I'd have to shoot at ISO 3200 or higher, or use some PS plug-in to add grain. Tri-X was my film of choice years ago -- often pushed a stop. Damnstraight it was grainy, and everyone was ok with that. It was Just The Way Things Were.

The only time I've cared about noise reduction was on noticing this strangely interesting effect: I had an ISO 3200 shot on-screen that I'd taken with a Nikon D3. You could see noise really only in pixel-peeping mode. But just to see what would happen, I gave the image some mild noise-reduction with the Topaz noise plug-in. Ok, so now the image is smoother but some of the sharpness is gone. What happens if I sharpen using the extremely slow (but extremely effective) sharpening tool FocusFixer? I'm going to get a possibly sharper image, but the noise will return, right? Wrong. What I ended up with was an image that was both smoother and sharper. The noise had not returned.

Hmmmmm...

[Mark D Segal]

I am somewhat puzzled for the significant concern voiced over a camera's inherent ability to shoot at higher ISO's with less noise, or the need to employ special noise reduction software, when the intended outcome is to produce a black-and-white print.  Isn't the net result of producing a black-and-white print with a higher level of RAW noise that merely(!) of having a print with an increased "grainy" effect, similar to various black-and-white films?  I always hav e the feeling in posing questions in this wonderful forum that I' asking really too basic questions, although no respondents to my queries have ever (yet) made me feel woefully (and inexcusably) ignorant.

Firstly, digital noise does not usually resemble film grain. Secondly, grain was a technical limitation of 19th and 20th century technology for most film. People learned to live with it, and even consider it a quality. Many still do today, and that's fine. With current digital technology we've moved beyond having to accept the fact of grain (inevitable for most films), and we have the choice of deciding how smooth or how grainy (and indeed with what character of grain) we want the image to look. That is an artistic decision we now have the freedom to make.

[walter.sk]

I am somewhat puzzled for the significant concern voiced over a camera's inherent ability to shoot at higher ISO's with less noise, or the need to employ special noise reduction software, when the intended outcome is to produce a black-and-white print.  Isn't the net result of producing a black-and-white print with a higher level of RAW noise that merely(!) of having a print with an increased "grainy" effect, similar to various black-and-white films?  I always hav e the feeling in posing questions in this wonderful forum that I' asking really too basic questions, although no respondents to my queries have ever (yet) made me feel woefully (and inexcusably) ignorant.

Unfortunately, digital noise does not look like film grain.  And, if you convert an RGB file to a black and white print, you still have two kinds of noise present.  The luminance noise is there, but so is a desaturated version of the color noise, and depending on how you convert to B&W, the color noise can be quite visible and contrasty.

In addition, depending on how you process the image, the following issues also make noise reduction desirable (although on a case by case basis):

1) Obviously, the size of the print.
2)  If you apply creative and/or output sharpening, any noise can be greatly emphasized.
3)  If you have to lighten shadow areas, the intensity of the noise is greatly increased.
4) If you have large solid areas such as sky, the noise may become too apparent.
5) If you do any formal portraits, going for a softer skin look, the noise could be objectionable.

These are just the issues that come to me off the top of my head.  In addition, if you like the use of film grain as an expressive technique, a program such as silver Efex Pro gives you so many ways of simulating grain that you should be satisfied, but the presence of noise on top of the grain can look quite distracting.

[Guillermo Luijk]

Firstly, digital noise does not usually resemble film grain.

I would say it never resembles film grain, because its physical origin is totally different. To understand the nature of digital noise, we cannot take our RAW converter and look at the images it produces, because the physical shape of noise will depend on the algorithm used by the software.

To understand why digital noise looks so awful, it is necessary to look or analyse the noise in the RAW data, i.e. the pure RAW information before it has been converted to an image. If we do so, we will realize sensor noise is the same as Photoshop's gaussian noise, i.e. a random deviation from the expected value on each pixel that has (almost) no correlation with neighbour pixels. Digital noise has no spatial structure at all, so it's nonsense even make any comparisions with film grain, which has a very well known and artistically appreciated shape. In the origin digital noise can always be considered one-pixel size, and the shape it turns into after RAW development only depends on the algorithm used, but it will never have a nice end.

In this figure we can see the RAW blue channel of 3 RAW files taken with different exposures (2EV apart). Their exposure was equalized in the end in order to make a comparision:

- The most exposed shot (+4EV) has very little noise.
- The medium shot (+2EV) displays a lot more noise, and its appearance is the same as Photoshop's gaussian noise (i.e. no shape, no grain, just random pixels here and there).
- The lowest exposure shot (0EV) displays even more noise, and because of the high noise amplitude with respect to signal many pixels got clipped to black producing that darker appearance.




But noise in the form of a random gaussian offset added individually on each pixel is something the photographer will never see, because he uses a RAW developer to create an image from the RAW file. It is in the process of RAW interpolation when this gaussian per-pixel noise becomes some textured colour and luminance noise, that will be in any case far from film grain in appearance.

Here is what a noisy RAW file looks like after being developed with two different algorithms:



They look slightly different (remember they come from the same RAW file), both very awful, and both far from pleasant film grain.

_______________

With regard to the impact of noise in B&W digital images, IMO cameras have enough resolution to make the appearance of digital noise after RAW development small enough not to worry about it in the print; it can even produce the illusion of enhanced accutance. And colour noise, which is actually very unpleasant and easy to notice in colour images, is not a problem here. So IMO noise in B&W digital images can be ignored.

It is good to keep noise low in shooting through a good exposure, because that will mean capturing more image detail. But once we have our RAW file, and detail cannot be improved, trying to eliminate noise in pp (at the risk of eliminating detail as well) could be considered not critical at all.

Regards

[JimAscher]

Wow.  You guys are great!  I guess I will just keep on as at present, shooting at as low ISO as feasible for the conditions and circumstances, and not worry too much about noise for black-and-white printing.  Many thanks to you all.

[Guillermo Luijk]

I will just keep on as at present, shooting at as low ISO as feasible for the conditions and circumstances

Be careful with staying at low ISOs because an underexposed RAW file at ISO100 will be noisier than a properly exposed RAW file at ISO200 (if you have Canon, you can nearly halve noise in the shadows just by pushing from ISO100 to ISO200 with the same aperture and shutter).

[JimAscher]

Be careful with staying at low ISOs because an underexposed RAW file at ISO100 will be noisier than a properly exposed RAW file at ISO200 (if you have Canon, you can nearly halve noise in the shadows just by pushing from ISO100 to ISO200 with the same aperture and shutter).

VERY interesting.  I have been shooting with my Panasonic Lumix's at 100 ISO to reduce noise, but with my Sony R-1 at 200.  (It doesn't go down to 100.)  So far I've not detected any under-exposure yet in my RAW files, but now I'll be somewhat cautious.  Many thanks.

[Lightsmith]

Most noise is chroma noise so it is going to much less evident in a B&W image. Test for yourself by shooting a picture at the highest camera setting you have available and again at 3/4 that ISO and again at 1/2 that level and compare the results.

Underexposure will generate a lot more noise in the finished image than having the camera amplify the sensor signals with a higher ISO setting. Better to shoot at ISO 3200 than at ISO 1600 if the later is underexposed by as little as 1/2EV. This too is easy to confirm with a dozen shots at -1 EV, 0.0EV and +1EV in several lighting situations with different subjects.

[Guillermo Luijk]

Better to shoot at ISO 3200 than at ISO 1600 if the later is underexposed by as little as 1/2EV. This too is easy to confirm with a dozen shots at -1 EV, 0.0EV and +1EV in several lighting situations with different subjects.

Shooting at ISO3200 provides an indistinguisable amount of noise from shooting at ISO1600 with the same aperture and shutter, no matter if using ISO1600 leaded to underexposure. The improvement is very apparent from ISO100 to ISO200, from ISO200 to ISO400,... (specially for Canon cameras, not so much for the Nikons), but becomes negligible in all cameras from ISO1600 in advance.

So for RAW shooters, ISO3200 and above are not very useful. In fact we can loose highlight information because of pushing ISO with no benefit in the shadows.

A real example, all original shots taken at the same aperture and shutter:


ISO3200 means losing highlights information with respect to staying at ISO1600...


...without any advantage in deep shadows noise.

Regards