Well, a couple of people know that I've been working on this, so I just wanted to share my findings to date.
I've been looking at how Anti-Aliasing math works, specifically how an "anti-alias filter" might effect things such as adjust>threshold or level-adjusted masks. It seems that true anti-aliasing comes from infinitely complex (pure math or vector) objects. Anti-Aliasing averages the edge weight in a pixel to the background, and rasterizes the result. What you get is the appearance of a smooth line.
Coming from a threshold or bitmap, right now I can't see (without vectorizing the edge) how an anti-alias filter would function. With 1-bit data, there could be no present data in the background space. The pixel is on, or it's off. So you'd have to average the edge through some sort of sketch filter or vectorization tool, and of course, that's inpercise. And clumsy and/or time-consuming.
The fractal guys have come up with an interesting solution though. A lot of the freeware fractal generators won't anti-alias either, so their method of anti-aliasing is actually rendering their fractals at four times (or more) their intended output size and reducing them in Photoshop to 25%. The necessary information between background and foreground becomes available with the additional pixel information. Obviously, the greater the increase/decrease, the more percise the edge anti-alias will be...
Thought I'd share... the only way I can see of making a true "anti-aliasing threshold filter" is by recreating the threshold filter but adding anti-alias functionality. For those of you interested in trying out the fractal method though, the information is available to you. For me, the job has gotten a little more difficult...
Any input or insight you guys have would be cool...
[Edited on 18-12-2003 by Phil_The_Rodent]





Reply With Quote
