[links-list] Re: font problem and fix

Josef 'Jupp' Schugt jupp at gmx.de
Sun Nov 10 16:35:34 PST 2002


* On 2002-11-10 23:19
* aludal <aludal at softhome.net> wrote:
> Look, when redesigning/retouching RusCyr and designing RusCyrBolds,
> I manually added hundreds of 'pseudoAA' grey pixels to pngs, or
> stretched/squeezeed these pngs to produce better smoothing ones, or
> delete stray ones. Without this retouch and without those grey
> pixels, I assure you, no 'resampled alpha mask' can correct an
> outline -- they looked ugly.

The problem seems to be a linguistic one.

The term 'antialias' only makes sense if you start with vector
graphics and end up with some pixel graphics or start with some
high-resolution graphics and end up with a low resolution one. I'll
only address the former one; you can always imagine the high-
resolution graphics as described by vector graphics.

Three cases are possible: A pixel of the target pixmap is covered
completely by the original graphics, is not covered at all or is
covered in part.

In the first (last) case the choice is obvious: That pixel must be
black (white). But what to do with the partly covered ones?

Plan A: If it is covered more (less) than 50% it becomes black
(white). This results in aliasing.

Plan B: Use an appropriate gray that reflects the coverage. This is
what antialiasing is all about.

Now we start with a black and white letter at the target resolution.
Such an image per definition has no aliasing artefacts. Artefacts
only exist if you have something for comparism.

The letter looks ugly because you have an idea of an ideal letter but
computers don't have an idea (pun intended), they do what they are
told to - how stupid that may be...

So you have two choices: Either tell the computer to do something
that results in a better looking character (like smoothing) or
communicate the idea to the computer (edit the file manually that
is).

Short: If you have a vector font the term 'antialiasing' makes sense,
if you have a pixmap one it doesn't.

HTH

(I did try to avoid references to sampling theorems and all that
mathematical stuff. IMHO Feynman was absolutely right in first
understanding a problem and then doing the mathematics and not the
other way around).

Josef 'Jupp' Schugt
-- 
Web: http://jupp.tux.nu/                           mailto:jupp at gmx.de
B.U.f.H. (Bastard user from hell): "Berserk Panic HOWTO"
while :; do cat $SHELL | wall ; done
-- 
Unsubscribe: send email to links-list-request at linuxfromscratch.org
and put unsubscribe in the subject header of the message



More information about the links-list mailing list