I just get so frustrated sometimes. I've read quite a few things about reducing halo when sharpening. They all say that halo is the enemy. They are wrong. The problem lies in the data and the algorithm - halo is a by-product of those two things.
They never directly address the problem of why halo is being produced in the first place - and I'm talking about a real hard look at why. When you look at halos from this perspective, better methods can be realized.
Bad: create halo and then try to reduce them.
Good: don't create halos to begin with.
So here is something that I've kind of mentioned in the past. I honestly don't know if anybody pulled it out and played with it, so I'm gonna pluck it out just in case you haven't.
- Random photograph you want to sharpen.
- Copy to new layer.
- On the copy, Filter > Noise > Median to taste.
- Invert
- Opacity 50%
You should now be looking at something very much like High Pass. However, halo should be close to non-existant. It will still be there depending on Median's radius and the data for a given part - but still greatly reduced or non-existant.
Now it's just a matter of your favorite method of Copy Merged and Paste. Then blending modes and what-not.
That was just adjusting the method. If you adjust the data before hand, you can get truly haloless sharpening. Get rid of the data that is causing the halo to begin with, and then fill it will data that does not cause halos. That's the crux of it, the basic idea.
Ugh.
play.fiddle.learn
posted 10-03-2007 08:28
