Sounds as if you've done a lot more work on this idea than Dan has--I would be really interested in the details. Is this a process you use on most images, then? (or was that just in 2003?)
I use it (in concert with Focus Magic) on all of my images.
The basic concept is straightforward. The image is converted to 16-bit, then to LAB. The L channel is duplicated as a mask, and a bell curve is run on the mask channel so that the shadows and highlights are not sharpened, but the midtones are. The action is paused at this point so that if you want to manually tweak the mask (For example, if you want to prevent the sky or a person's face from being sharpened, you can airbrush that part of the mask to black and it will not be touched). Then the mask channel is loaded as a selection, and successive rounds of USM are done on the L channel with progressively smaller radius and higher amount values. After that, the mask is deleted, the selection is removed, and the Shadow/Highlight tool is run on the L channel to bring out detail in the shadows and highlights that may have been de-emphasized by the previous rounds of masked USM. Then the image is converted back to RGB.
The theory of operation has some similarities and differences compared to PK Sharpener. Like PK, it focuses on the midtones, but unlike PK, it doesn't really separate between capture, creative and output sharpening. If you think of an image in audio terms, one could equate the subwoofer volume level with what is commonly called "local contrast enhancement" (usually USM radius 100 or larger), fine pixel-level sharpness (USM radius 1 or less) as your high treble adjustment, and intermediate radius settings as frequency bands falling in between the extremes. When sharpening, one needs to appropriately adjust the image throughout the range of frequencies, in the same way that adjusting only the low bass or high treble on a stereo equalizer is not usually going to result in very good sound. IMO, this is a major flaw in most sharpening schemes; they only adjust either the "high treble" (USM radius 1 or less) or "low bass" (USM radius 50 or greater), and usually ignore the "mid-range" of radius values.
PK sharpener takes a 3-pass "equalization" approach comparable to having an equalizer in your CD player (capture sharpening), a second in your amplifier (creative sharpening), and a third in your speakers (output sharpening). There's nothing inherently wrong with this approach; having the equivalent of "EQ presets" for various input and output devices has a lot of merit, especially when making multiple versions of an image file for different output devices. But while certainly useful in many situations, "EQing" your image in 3 passes isn't strictly necessary. If you're not constantly changing your CD player and speakers, a single equalizer in the path between player and speakers is sufficient to correctly adjust the levels of the various frequency ranges to result in a properly-balanced output from the speaker.
That's more or less what my action is doing, functioning as a "single EQ" to properly balance the various spatial frequency ranges in an image in one step. When running the action, each round of USM is comparable to adjusting one frequency slider on a graphic equalizer, starting with low bass and working up to the highest treble range. It's not necessarily as fancy as PK sharpener, or always as convenient to use, but can deliver comparable results when used properly. The most important thing to remember is that since there are several rounds of sharpening, it's OK (and in fact necessary) to save something for the next round(s) of sharpening to do. Adjust the Amount slider until you start seeing haloing or pushing the boundaries of good taste, and then back it off some. As the action runs, you should see the image detail and micro-contrast gradually increase with each round of sharpening until you have the final finished product. If you're sharpening for print, you'll need to know how "crunchy" to make the image look on-screen (this will vary depending on printer and paper type, and is where PK Sharpener can save you some time).
In my workflow, I use Focus Magic for capture sharpening, as it uses a deconvolution "unblurring" algorithm that that does a much better job of undoing the effects of the anti-aliasing filter than simple USM or similar sharpening methods. I usually use Radius 2, 25-50%, followed by Radius 1, 25-50%, exact settings depending on which lens I'm using and aperture and such. Then I run my action to do creative and output sharpening combined.
