Two-Pass Sharpening
Several years ago, I started playing with the notion of doing two separate sharpening passes. The first, pre-sharpening pass aims simply to restore the sharpness lost when turning the image into pixels, whether by scanning or direct digital capture. The second pass is then tailored to the particular printing process that will be used.

Separating the two passes lets you address two very different issues separately -- the sharpness the image itself needs, and the sharpening the printing process requires -- and when done carefully this approach works very well indeed. Like most things in digital imaging, however, when done carelessly, two-pass sharpening can make a pig's ear of your image.

First Pass
As your first pass, if you simply apply a gentle sharpening to the entire image with the Unsharp Mask filter, your image will look great -- but only until you try to sharpen it again for output. At that point, you'll quickly find that textured areas start to look crunchy, and that noise pixels pop out like stars behind the new moon.

A much better approach is to create an edge mask -- a mask that selects only the high-contrast edges, leaving the rest of the image protected -- and to apply modest pre-sharpening through the mask. In extreme cases, when the image is very noisy, you can even invert the edge mask and apply some kind of gentle blurring or smoothing, which lets you not only avoid sharpening the noise in the image but also actually suppress some of it; I usually use the Dust and Scratches filter with as low a radius and as high a threshold setting as possible, but Median and Gaussian Blur also work for this. In most cases, though, a slight boost to the contrast along edges is all that's required.

Making an Edge Mask
Adobe Photoshop offers many tools for making masks. Here's one technique I've found useful and reliable. Figure 2 shows the image before pre-sharpening.

Figure 2: Before pre-sharpening

I start by duplicating the image (as a matter of convenience when working with an 8-bit-per-channel image and as a matter of necessity with high-bit images, because most of the tools used in this technique aren't available when editing high-bit images). Then I convert to grayscale, using whichever conversion method provides the best contrast along the edges that need sharpening. Which method works best for you will depend on the image you're working with, and this topic alone could fill an entire article. Often, though, simply duplicating one of the individual color channels using the Channels palette works well. In other cases, I may use the Channel Mixer or Calculations to blend two or more color channels into a grayscale version. If you're starting with a high-bit image, you also need to downsample the duplicate to 8 bits per channel by choosing 8 Bits/channel from the Mode submenu on the Image menu. In this case, I simply duplicated the Green channel into a new document, shown in Figure 3.

Figure 3: Image converted to grayscale, in this case by duplicating the green channel into a new document.

The next task is to isolate the edges. The easiest way to do so is to run the appropriately named Find Edges filter, which you can find on the Stylize submenu in the Filter menu. This produces the result shown in Figure 4.

Figure 4: Here we isolated the images using the Find Edges filters.

This does a reasonable job of finding the high-contrast edges in the image, but it also picks up some noise (in this case, the noise is film grain from ISO 400 color negative, but in other images it may pick up scanner noise, or a noisy blue channel in a digital capture). So I need to tweak the mask some before I can apply it to the image.