home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.fonts:3746 gnu.misc.discuss:4091
- Path: sparky!uunet!noc.near.net!hri.com!spool.mu.edu!agate!doc.ic.ac.uk!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.92Dec22183541@decb.aber.ac.uk>
- Date: 22 Dec 92 18:35:41 GMT
- References: <PCG.92Dec13195105@aberdb.aber.ac.uk> <1992Dec17.193208.11337@dircon.co.uk>
- Sender: news@aber.ac.uk (USENET news service)
- Reply-To: pcg@aber.ac.uk (Piercarlo Grandi)
- Organization: Prifysgol Cymru, Aberystwyth
- Lines: 89
- In-Reply-To: uad1077@dircon.co.uk's message of 17 Dec 92 19: 32:08 GMT
- Nntp-Posting-Host: decb.aber.ac.uk
-
- On 17 Dec 92 19:32:08 GMT, uad1077@dircon.co.uk (Ian Kemmish) said:
-
- uad1077> The grid-fitting in TrueType operates purely on the geometry of
- uad1077> the outline and doesn't care (and can't find out) how many
- uad1077> colours your `pixels' can represent.
-
- Yea, that's what I'd call a problem. Given that color screens are
- fuzzier than mono screens, and that for some silly reason color screens
- are prevalent, it seems strnage not to make use of the ability to
- antialias. I see that antialiasing is even more sensible on a grayscale
- screen, as its very crispness makes text even more readable.
-
- uad1077> You may be being misled by thinking in terms of supersampling
- uad1077> as the only way of achieving anti-aliased output from
- uad1077> scan-conversion,
-
- No, no, I was thinking of that as *a* way, and a cheap one too. My
- posting was about *two* independent observations:
-
- 1) How strange that antialiasing is so unpopular, when it is so nice,
- and can be so effectively done by shrinking printer resolution bitmaps
- onto the screen.
-
- 2) Storing "ideal" character shapes as (approximate) high resolution
- bitmaps, and then shrinking them to lower solution ones, with some
- opportune optically wise filtering algorithm, looks like a faster and
- otherwise equivalent technology to storing "ideal" character shapes as
- (approximate) outlines to be fleshed out, with correction by hints.
-
- The link between shrinking and antiliasing is that shrinking from a
- bitmap to a pixmap gives excellent results even with fairly simple
- algorithms and is probably way less computationally expensive than
- outlining to a pixmap.
-
- uad1077> and further, thinking that you then have to apply the hints on
- uad1077> the granularity of the supersampled grod, not the original (or
- uad1077> final, depending on your viewpoint!) grid. Neither of these is
- uad1077> necessarily true.
-
- Agreed; in effect I did not quite intend to say this. What I wanted to
- mean is that probably shrinking can be done with some algorithm that
- gives the same results as hinting; hinting after all is just a way to
- compensate for the lack of detail in the final low resolution bitmap by
- using optical tricks.
-
- uad1077> The only real survey I saw done on anti-aliased text (John
- uad1077> Amanatides? I'm not sure - I'll go and check)
-
- Yes please! A reference I got by e-mail (thanks!) is to the work on this
- done for a troff preview by a person at AT&T. Appanrelty he found out
- that given the way the eye perceives grids it is possible to build a
- filter that shrinks to a bitmap that looks optically closer to the effect
- of the original than more trivial ones (e.g. averaging), and gives much
- the same effect as hints. I'm going to pursue that reference.
-
- uad1077> showed that prolonged viewing of anti-aliased text induced more
- uad1077> eyestrain than decently tuned bilevel text.
-
- Ah, this is another matter entirely. If so, it's unfortunate.
-
- However I am surprised; I would be less surprised if the comparison was
- between bitmap fonts on a B&W screen and pixmap/antialiased fonts on a
- color screen, as color screens are noticeably fuzzier than bitmap
- screens.
-
- uad1077> I can generate anti-aliased PostScript output (by
- uad1077> super-sampling) and look at it on the Sun screen, and while I
- uad1077> can recognise words at only a few pixels tall, I certainly
- uad1077> wouldn't like to debug with that text - I'd rather spend a bit
- uad1077> more on a half-decent size screen and use good 15-pixel bilevel
- uad1077> text.
-
- To me it is interesting to compare, given the same hw, two sw solutions
- (outlining+hinting vs. large bitmaps+shrinking), rather than assuming
- that it is possible to get better hw for one and not for the other.
-
- If one can afford a crystal clear > 17" screen with a 120 DPI resolution
- and hand tuned fonts for outlining+hinting, and not for
- bitmaps+shrinking, then too bad for the latter.
-
- But if we are comparing two identical sets of hw, then maybe starting
- with 600DPI bitmaps, and either antialiasing them to 100DPI for
- previewing or filtering them to 300DPI for printing might be a better
- (faster, less proprietary, maybe even better looking results for both
- bitmap and pixmap case) than outlining+hinting.
- --
- 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
-