[Definition] Unsharp Masking

[john_opitz]

And the reason why the red(when in rgb) channel. Since its lighter in faces, so the sharpening that takes place. It will not sharpen the detail in the faces. While sharpening other parts of the picture(hair,clothing, eyes,ect...ect...)

You can also sharpen with layer masks,actions,custom filter, but when using these things. You have to know whats going on and how these things affect the image when sharpening.
Like some plug-ins for sharpening. Their not bad. If one does not understand the sharpening process. But you can do the same in P.S. and better when you know the sharpening process in p.s., and the techniques that go beyond the typical all channel sharpen.

I feel its very important to look in your channels, when sharpening. Sometimes one doesn't have that kind of time though when working on a lot of images a day. So that's where actions, plug-ins and an all channel sharpen comes in handy for sharpening.
It's like in martial arts. You have to know when to use your feet instead of your hands. Execpt in kempo. That form is more, using your hands than your feet.

[john_opitz]

<<Therefore, I would have thought that the green, rather than red channel would be the one to filter.>>

As Steven just stated...... That depends on the channel. Let me explain why I sharpen as I stated. I deal with a lot of faces. So the red channel is best for me. But one thing.... You have to look in your channels to see what going on in them, like noise,artifacts,ect...ect.. which you don't want to sharpen.
Btw. Steven has a great web page with links as well that go into this. Check it out. Btw. I don't have a web page for P.S. .Just don't seem to have the time. Anyway it would take me forever to make(type) one.
John

[Stephen M]

Chuck, it depends on whether the R or G is best in RGB (and perhaps even the B in some cases) - if the image is of one major hue only. In general for an image with a good range of luminosity and colour, then the G is often the best choice for a luminosity blend sharpen.

I personally prefer a luminosity blend or fade rather than converting to LAB or HSB or anything. Where possible I avoid mode changes or do these in a dupe for targeted reblending back into the original in luminosity or colour blend modes.

As John mentioned, when doing separate channel sharpening - the luminosity blend is important to stop colour halo shifts (just as luminance sharpening is important when sharpening the colour composite channel).

More links on sharpening can be found at my links to more links:

http://members.ozemail.com.au/~binaryfx/PSTV_links.html#S

Regards,

Stephen Marsh.

[Sam]

Thanks for that VERY useful tip, John. I'm off to try it out right now!!

[john_opitz]

When sharpening. I do a channel,by channel sharpening.
When in CMYK, I sharpen the black and cyan channels only. When in RGB, I sharpen the red channel only. When in LAB, the L channel only. And always fading the effect of the sharpening to Luminosity mode. No color shifts this way.

Just so happens, here's an article on this method.

Sharpening

[Doug Nelson]

"Unsharp Masking" sounds like a bit of an oxymoron. The term goes back to pre-digital prepress days when they'd make an image appear sharper by sandwiching it with a slightly blurred (unsharp) copy of the same image.

The effect then was the same as the digital effect now: the intersecting areas of light and dark in the image have an increased local contrast, which gives the appearance of increasing sharpness.

You can see the process more clearly if you grossly overapply USM to an image. The intersection contrast will be exagerated to a degree where they'll actually bloom over into each other, causing the "halo" effect so typical of oversharpened images.

Calculated the actual degree of USM to apply is actually rather complicated, since you need to take into consideration the qualities of the output device, the size of the image, and even the typical viewing distance.

It's important to remember that you're only giving an impression of increased sharpness, not actually increasing the sharpness, and that you're technically degrading the image. Also, since there are so many variables, you'll probably want to make separate printing copies of your image clearly labeled as to destination and apply USM specifically for that destination. This allows your original to remain unsharpened so you can make another copy should you need to prepare it for a different destination.