If you find yourself shooting in low light situations then digital noise is likely to bedevil your results. That is to say, as you push the ISO higher to make the sensor more sensitive to light, you will get grain and pixel distortion that can ruin a picture.
I find myself shooting a fair bit of low light photography and the best thing you can do to avoid noise is to have fast glass (lenses with a wide aperture, like below f/2) and a camera body that handles higher ISO well. However, digital noise will still be an issue. Happily, there are some great software applications out there that will logarithmically analyze your photo and reduce the noise by extrapolating data from “good” pixels and substituting that new data for the bad or noisy portions of the image.
In the past I used a product called Noise Ninja. Noise Ninja is a plug-in that works with Aperture, Photoshop and Lightroom. It did a pretty good job but its main failing was that it applied these changes globally which means that it made judgements about the noise in an image then applied the “fix” to the whole image. The downside of this was that parts of the photo that were not noisy ended up getting the same treatment as the parts that did exhibit noise. Many times this meant that detail was lost. In particular, Noise Ninja has the effect of making people look like figures in a wax museum.
However, Nik Software’s Dfine 2.0 (also a plug-in for the above mentioned software) does not make noise correction globally. It analyzes the photo in question and applies noise reduction in different amount to different parts of the image. This makes a world of difference, pardon the pun.
The shots here were of a recent show at Red Star (downtown Baton Rouge’s finest drinking establishment). Graham Colton was playing the room so I brought out the camera and got a few shots. One thing that I am always amazed about with low light photos is that while the 5D MKII is pretty lousy at pulling focus in low-light, low-contrast situations, once you figure out how to get the subject into focus, the high performance of the Canon sensor can deliver serviceable images up to ISO 3200 and probably even beyond.
What you are seeing here is one shot with a dark area (above the singers right shoulder and under the painting on the wall) cropped and blown up. One version is before I applied noise reduction through Dfine and one is after it was applied. I have also posted a close crop of the singers face so you can see that detail is still maintained at an acceptable level. Shots at ISO 3200 (with present technology) are always going to be a bit coarse, but you can see plenty of detail and, frankly, it is pretty incredible when you consider how little light was really present at this show.
Another thing that made shooting this show a bit easier than some other low-light performances was that we use standard, incandescent bulbs in our stage lights. This provides a pretty even color temperature unlike LED floods which will just give you fits when it comes to color correction.
Within Dfine’s architecture there are options for manually applying noise reduction wherein you tell the program what areas to sample but it generally does a good job of finding both the contrast and color noise and separating the areas out so it can work on them. From time to time I do find myself having to direct the program to the trouble spots, but these shots were all edited using the default settings. For the sake of comparison I am posting a version of the shot that was edited with Noise Ninja (also under its default settings.) You be the judge… but not until after you have clicked on the various images and seem them in a larger version. I think it’s pretty telling.