«
»

Color Noise

I like grain. I’m hardly unique in this — grain has a long and noble history in photography. One of the awesome things about shooting film is the weirdly satisfying look and feel of film grain.

Cigarettes and Coffee

Nose to the Water

Good-looking grain makes it possible to shoot at very high speeds and still get usable results for many applications, and in many cases, it’s a critical part of the image.

Digital doesn’t have grain, it has “noise.” In addition to sounding much less awesome than grain, noise also, sadly, does not look as awesome, mainly because it’s garishly colored. I don’t have an example of this, so I won’t post one, but I’m sure you know what I mean.

There are two obvious approaches here. The most obvious one is noise reduction. I find that I use less and less of this as time goes by. Sometimes it’s really needed, and sometimes it really works, but sadly, those times don’t necessarily coincide, and when noise reduction goes wrong, the resulting muddy softness is even more irritating than noise. I could probably invest a lot of time and energy in getting better at noise reduction and sharpening, and at some point I will have to do that, but in the time, I’m trying to cut back on it.

The second obvious approach, which I use maybe a little too much, is black and white conversion. Black and white conversion of digital noise doesn’t result in film-like grain, quite, but for many applications it’s fairly close, and fairly pleasant, even when shooting at ISOs that are far higher than what’s really advisable.

Predawn

This approach works well in a lot of situations, especially in low-light conditions where colors often become boring and/or ugly, anyway. And, indeed, in any situation in which the important visual information is to do with contrast and tonality rather than color (and this is many, if not most, images), black and white is probably a better choice, anyway.

But, sometimes color is important. For shots in which you have objects with similar tones but very different hues, a black and white rendition will be quite boring, and quite possibly misleading. And if one of the goals of an image is to accurately depict reality, and the subject is known by the audience to be colorful, a black and white image will often fail horribly.

So, what I do now to deal with color noise in digital files, when I want to preserve color in a scene, is to desaturate the noise.

Olympus XA

I do this using Capture NX, which is a program of odd strengths and weaknesses — the strength in this case being its inclusion of Nik Software’s U-Point technology, which is lately being marketed as a Photoshop/Aperture plugin called Viveza.

The points, called “Control Points” in NX, are little dots placed in areas of the image, each sporting an area of sliders for brightness, contrast, saturation, RGB, etc. You use the sliders to tell NX what to do to parts of an image, within a certain radius, that match the part you dropped the point in. A few more points placed in different areas, either to make other changes or to protect from changes by surrounding points, and you’ve (hopefully) got your image just how you want it. It’s a profoundly better way to do local edits than layer masking, and by comparison, burning and dodging in the darkroom is a tremendous burden.

20080421-008

The trick is to place a control point in an area of intense noise, and then completely or nearly completely desaturation it. Then place a second control point in a nearby area of the same overall color, but without so much noise. This will result in some overall desaturation, but mainly it will reduce the color noise, leaving you with a grainy speckling, but none of those annoying red, green, etc. dots — and the overall effect is much more pleasing.

Tags: , ,

Comments are closed.