Re: [Devel] Adding options to the PS Hinter

[Vadim Plessky]

On Sunday 07 July 2002 12:35 pm, David Turner wrote:
|  Keith Packard wrote:
[...]
|  >
|  > A harder problem would be to retain the angular relationship between
|  > diagonoal elements in glyphs; I've seen the autohinter do weird things
|  > with ¥ in some fonts.
|
|  I've been thinking about it for a long time, however, the problem is not
|  necessarily trivial. Here are a few points to consider:
|
|    - I believe it's a bad idea to work on diagonal auto-hinting until the
|      hinting of normal stems isn't completely correct. For now, there are
|      still problems with serifed fonts, and these introduce distortions
| that are much more important, and which also "nuke" the outline so
| seriously that attempting to hint diagonal isn't going to work or be
| noticeable.
|
|      We could start working on hinting diagonals of sans-serif fonts, but
|      there are chances that the algorithm doesn't work well with serifed
|      fonts and would need re-working..
|

Hi David!

After 2 threads devoted to rendering quality of PS fonts (<freetype> mailing 
list)
[Freetype] FT 2.0.9 rendering for some glyphs
from 16.04.2002
[Freetype] Test: Native PS T1 hints vs. PS Auto-Hinter on TrueType outlines, 
from 7.05.2002.

I did some further investigation with Serif'ed fonts.
And, as I found, there are some fonts which behave much better with PS Hinter 
than Times New Roman, Nimbus, Newton (mentioned in my message) and some 
others.
As a reference, I attach small screenshot for one Serif font ('Dutch'), which 
is some kind of clone for Times/Times New Roman.
As you can see from screenshot, rendering for 11pt is close to perfect, only 
excpetions is glyph W, and may be M N Y V need some small enhancement.
Screenshot was rendered using FT 2.0.9.
Font is PS Type1 font, manually hinted.
Comments from your side are highly appreciated!

There are more artefacts at 12pt rendering.
Latin A, K, M, N, O, Q, R, V, W, X, Y, v, w, x, y, 
Cyrillic Э, Ю, У, а, б, м, х (in addition to glyps similar to Latin)
have some artefatcs, but theyare *minor*,  comparing to problems of rendering 
Nimbus or Times New Roman.

So, I believe, there are some issues with Times New Roman and Nimbus Roman 
fonts theirself. Probbaly, it's some kind of combinationof stem widths, 
height and hinting information.

|
|    - hinting straight diagonals isn't that difficult once the glyph has
| been properly analyzed. On the other hand, hinting diagonal curves is much,
| much, more problematic. there are lots of "special" cases to handle.
|
|    - sometimes, the diagonals are, like the ones in ¥ are split in several
|      segments, and this makes the analysis a lot harder, and the alignment
|      subtler.
|
|    - finally, it'd be nice if the auto-hinting algorithm could still be
|      performed relatively quickly. Deep diagonal analysis requires more
|      CPU and I'm not confident that the performance of the hinter will
|      still be acceptable. Same goes for counter-hints analysis (a really
|      tricky subject too). An alternate solution would be to pre-compute
|      glyph hints in an external file (that one could FT_Attach) to avoid
|      re-computing them everytime though..

Ok, it's clear that Auto-Hinter needs extra work and solution like external 
hint storage.
But pre-hinted PS Type1 fonts don't need it, I hope. :-)
Can grayscale, anti-aliased rendering of PS Type1 fonts fixed first?..

|
|  Hope this helps,
|
|  - David Turner
|  - The FreeType Project  (www.freetype.org)
|
|
|  _______________________________________________
|  Devel mailing list
|  address@hidden
|  http://www.freetype.org/mailman/listinfo/devel

-- 

Vadim Plessky
http://kde2.newmail.ru  (English)
33 Window Decorations and 6 Widget Styles for KDE
http://kde2.newmail.ru/kde_themes.html
KDE mini-Themes
http://kde2.newmail.ru/themes/

dutch-11pt-12pt.png
Description: PNG image