Re: Yet another anti-aliased KDE screenshot
[ PGP not available, raw data follows ]
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On Tuesday 30 October 2001 05:48, Daniel Robert Franklin wrote:
> > So, in short, if you actually want to see the font, as opposed to a bunch
> > of artifacts, you need to use anti-aliasing. Don't mistake smoothness for
> > blurring.
>
> That was some nice explanation. Anti-aliasing is something you ought to
> integrate in any display system, or you will have jagged edges. Thus, some of
> the hype associated with Mac OS X's Carbon.
Definitely. OSX takes it one step further though, although I don't think it
takes it far enough. My personal preference would be the total elimination
of the "pixel" as a concept for user-interface design, rather to have
everything in terms of vectors, geometric lines etc, then use a simple
rendering engine to sample that and turn it into pixels. Then you would be
able to do arbitrary scaling, translation and rotation with the aid of
hardware acceleration (much like the user-interface on the excellent QuArK
Quake map editor). You could use scalable dimensionally-normalised eps files
or some such thing for icons, buttons etc. I don't know if the WIMP paradigm
would still work in that environment, but it certainly would be interesting
to find out, or to develop something better :)
>
> I'd always thought anti-aliasing was simply resampling, tho'.
Well, AA comes into that too - if you have an image with 600 dpi resolution
and you want to display it on a monitor with 100 dpi resolution you need to
downsample it. But this is not as simple as taking every 6th pixel (though
that is the way it is sometimes done). You need to low-pass filter it first,
or you will see artifacts. The effect is similar to sampling an analog
signal. Accurately changing the sample rate to something which is not 1/N (N
being an integer) (e.g. from 600 to 83 dpi) requires upsampling and then
downsampling - but you still need to filter.
> Which means there is actually some blending but not blurring, most
> algorithms are careful enough not to degrade image quality. On the other
> hand, the end result varies since there are different algorithms.
Yes. The precise digital filter used can be of many different designs, but
all are imperfect (a digital filter would need to be infinitely long for it
to be perfect). You will always get some aliasing but it should be much less
than with no filter (hopefully below the threshold of perception).
Forgive me for rabbiting on, as a DSP-obsessed electrical engineer, it's a
subject which is close to my heart :)
- Daniel
--
******************************************************************************
* Daniel Franklin - Postgraduate student in Electrical Engineering
* University of Wollongong, NSW, Australia * d.franklin@ieee.org
******************************************************************************
Reply to: