home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.fonts:3822 gnu.misc.discuss:4124
- Path: sparky!uunet!pipex!warwick!uknet!gdt!aber!fronta.aber.ac.uk!pcg
- From: pcg@aber.ac.uk (Piercarlo Grandi)
- Newsgroups: comp.fonts,gnu.misc.discuss
- Subject: Re: "Hinting" using antialiasing? A GNUish alternative...
- Message-ID: <PCG.92Dec31122838@decb.aber.ac.uk>
- Date: 31 Dec 92 12:28:38 GMT
- References: <PCG.92Dec27213908@decb.aber.ac.uk>
- <1992Dec27.225347.28173@pasteur.Berkeley.EDU>
- <PCG.92Dec29154125@decb.aber.ac.uk>
- <1992Dec29.193853.17902@pasteur.Berkeley.EDU>
- Sender: news@aber.ac.uk (USENET news service)
- Reply-To: pcg@aber.ac.uk (Piercarlo Grandi)
- Organization: Prifysgol Cymru, Aberystwyth
- Lines: 126
- In-Reply-To: johnm@cory.Berkeley.EDU's message of 29 Dec 92 19: 38:53 GMT
- Nntp-Posting-Host: decb.aber.ac.uk
-
-
- A point of order: I will eventually post a summary of the e-mail
- suggestions I have received. If any of my correspondents objects, please
- let me know; after all I did not say explicitly that I might do a
- summary. Thanks!
-
- On 29 Dec 92 19:38:53 GMT, johnm@cory.Berkeley.EDU (John D. Mitchell) said:
-
- johnm> The Nimbus font driver for GEOS is the fastest rasterizer
- johnm> implementation I've ever heard of, seen, and/or used for that
- johnm> class of hardware.
-
- Some info about Nimbus? Some details about the rasterizer? :-)
-
- pcg> For example I don't like the few seconds of latency that existing
- pcg> renderers make me wait while the rasterize the fonts contained in an
- pcg> imported document.
-
- johnm> I'd say that it's most likely a matter of poor
- johnm> design/implementation of the ImpEx code rather than (just) a
- johnm> problem of the rasterizing of a bunch of fonts/characters.
-
- No, because the delay does not appear when prebuilt bitmap fonts are
- used...
-
- johnm> Perhaps the systems that you're using are doing something silly
- johnm> like building out all of the characters in each font that's used?
-
- Neither, because it is easy to see a small pause as a previously unseen
- character is rasterized in each of the first few paragraphs; after the
- first few paragraphs virtually all the letters are done, so it's quick.
-
- johnm> Fixed point arithmetic, when intelligently applied, is a great
- johnm> thing. :-? :-)
-
- Quite so, but not so easy; I imagine from your smileys that you have had
- your little time with it.
-
- johnm> Seriously though, I don't see any way of producing the kind of
- johnm> quality output that you desire without non-local contextual
- johnm> information and that must dramatically increase you resource
- johnm> needs.
-
- Bah, I can only hope some sucker (maybe myself :->) sees to it that non
- local information filtering be not that slow. For example, instead of
- averaging a 4x4 cell to 1x1, one could do with 8x8 to 2x2; or indeed one
- could go for the really expensive and do 8x8 to 1x1. But it may be still
- be fast enough: I still think that simple bitmap sweeps might be more
- suited to current hardware than drawing a curve and then filling.
-
- pcg> I think it would be more general because I guess it could be
- pcg> patterned after the rather universal properties of human vision;
-
- johnm> The same argument as above. Any typeface designer worth anything
- johnm> must take into account the biases and pecularities of human
- johnm> vision.
-
- pcg> As agreed above, but they do so *manually* (and "unconsciously":
- pcg> they basically assume that what looks right to them looks right to
- pcg> everybody). I think instead that it can be done, with tolerably
- pcg> satisfying results, algorithmically.
-
- johnm> I think this moves off into discussions about expert sytems and
- johnm> other AI shme.
-
- Not really; I think more straight image processing; there is an amazing
- wealth of work that has been done on bitmap/pixmap filtering.
-
- johnm> I do believe that many of the 'standard' tricks and rules of the
- johnm> human vision trade can be codified into algorithmic 'filters' but
- johnm> I don't see how that will be 'good enough'.
-
- The only evidence I can offer are that automatically applied hints, as
- in FontoGrapher, are often 'good enough'; also METAFONT, see below.
-
- johnm> As they say, knowing when to break rules is at least as important
- johnm> as the rules themselves.
-
- Ah, but then this amounts to adjusting by hand the output bitmap for the
- very most elegant output; and for this I think that just doing that and
- recording the differences from the scaled one would do.
-
- pcg> I like more what is done in METAFONT; there adaptation to the
- pcg> resolution and quirks of the intended output device is done by
- pcg> adjusting various parameters, like the shape of the "nib". This is
- pcg> virtually equivalent to choosing a different filtering algorithm
- pcg> (the "nib" is a filter!).
-
- johnm> Hmm... Sounds interesting. So instead of a single 'virtual' pen
- johnm> you can have a multitude of drawing tools?
-
- Indeed; when the countour of the outline is drawn, this is done with a
- shaped "nib", and the size and characteristics of the nib can be
- adjusted; for example METAFONT will rasterize to different bitmaps
- depending on whether the output device is a write white or write black
- laser (something no outline rasterizer that I know can be parameterized
- about, except at times those embedded in the printer itself, that get
- hand-tuned and hardwired, but not always, to its characteristics -- even
- with Adobe, two printers with different technologies are more liable to
- give different results with Type 1 fonts than with METAFONT ones). An
- e-mail correspondent (Brian Bartolemew) observed:
-
- Brian> I've long suspected metafont knows more about low resolution (~90
- Brian> dpi) than it lets on - witness the comments-in-passing about
- Brian> using diamond- shaped nibs instead of elliptical ones because
- Brian> they produce fewer artifacts at lower resolutions.
-
- Indeed even CM fonts, which have fairly fine details, come out quite
- well at screen resolutions, without any hints whatsoever.
-
- johnm> P.S. I'm certainly for more/better 'free' scalable font
- johnm> technology and building this stuff on top of filters may or may
- johnm> not prove to be a good thing. As an issue more relevant to the
- johnm> gnu side of this discussion, I worry that the 'filters' being
- johnm> 'programs' would thus be much easier to control (in the
- johnm> copyright/patent sense).
-
- Ah yes, but virtually all the image processing techniques around are
- well known art, and thus free. Not that this has stopped anybody ever;
- as the comp.compression FAQ reports, RLE encoding has been patented in
- 1986...
-
- --
- Piercarlo Grandi, Dept of CS, PC/UW@Aberystwyth <pcg@aber.ac.uk>
- E l'italiano cantava, cantava. E le sue disperate invocazioni giunsero
- alle orecchie del suo divino protettore, il dio della barzelletta
-