Noise is an inherent feature of digital images. Digital images are composed from millions of photo receptors. Not every one of those millions of photo receptors responds exactly the same to the same level of light. Add to that, the fact that the reaction of digital image chips is not perfectly linear, which is a fancy way of saying that they are more sensitive to bright light than they are to shadows.

ISO 200, Low Light, Canon 1Ds MkII

Noise tends to fall into two broad categories: luminosity noise and chroma noise. If you shoot at high ISO with a digital camera, you will likely see lots of colored speckles. That's chroma noise. Ever look at shadows or a gradient, like the sky or water, and see what looks like splotches or clumps or muddiness? That's luminosity noise.

Shadows at Actual Pixel size. Luminosity noise.

There are a number of techniques for removing noise. Some basic Phoitoshop tricks, like converting to L*a*b and running Dust and Scratches filter on the "a" and "b" channels will help. My preference is to use Neat Image Pro+. It is a sophisticated program for removing noise than can run standalone or as a Photoshop plug-in. There are other noise removal programs, too. Noise Ninja is another popular program for removing noise.

You need to be careful when you run a noise removal program. The idea is to remove noise without softening the image.

Noise is most noticeable in gradients and areas of color without much detail. That's why it often shows up in shadows and bright blue skies.

Sharpening, on the other hand, is most obvious when it is applied to edges. There is no point in sharpening a bright blue sky. The clouds, maybe. But large swatches of blue sky gain no benefit from sharpening.

The idea, then, is to keep most of the sharpening effort on the edges and most of the noise removal away from the edges.

Experienced digital photographers use an edge mask to keep sharpening focused on the edges. The Capture sharpening actions in the TLR Sharpening Toolkit use edge masks. The toolkit also includes actions for generating edge masks, even if you use another tool for sharpening, like Focal Blade.

If you invert an edge mask, you have a surface mask. This is perfect for noise removal. (The TLR Sharpening Toolkit also includes actions for generating surface masks.)

The workflow with this image started in Adobe Camera RAW.

A round of capture sharpening was applied with the TLR Sharpening Toolkit. USM settings were 370, 0.7, 2. The image size here is rather small to appreciate the settings. Sharepning was done with the image zoomed to Print Size and the USM window set to 800% to watch the sharpening halos.

I apply sharpening, noise removal, and other edits to layers. If the top layer is a pixel layer, all you need to do is duplicate it. Otherwise, you need to merge the underlying layers into a new layer and apply noise removal to that.

Neat Image Pro+ did a fine job of noise removal. I stronly endorse Neat Image Pro+ for noise removal. It is a sophisticated tool that does its job well, and Vlad does a great job of customer service.

Luminosity noise removed with Neat Image Pro+.

To keep the noise away from the edges, where it could perceptibly soften the image, I bopped over to the Channels palette, duplicated the edge mask, and then inverted it (ctrl/cmd + i). I made the new channel into a selection and returned to the Layers palette, where I added it to the layer as a Reveal Selection layer mask (via the Layer menu).

Feel free to download my TLR Sharpening Toolkit. The mask actions make it quick and easy to add a surface mask to your image for improved noise removal that doesn't work against your sharpening efforts.

Cheers!