Re: [ft-devel] ttfautohint

[Werner LEMBERG]

> Perhaps some such "slowish but high quality" optimizations could be
> added to the autohinter itself, as an optional mode for platforms
> powerful enough handle it well (presumably, deselectable at
> compile-time, for embedded platforms etc).

Yes.  There are three kinds of improvements:

  (a) analysis of glyph shapes

  (b) better rasterization

  (c) a mixture of (a) and (b)

An example of (a) is given by these two articles:

  Perceptually-Tuned Grayscale Characters based on Parametrisable
  Component Fonts
  (http://diwww.epfl.ch/w3lsp/publications/typography/ptgcbopcf.html)

  Parameterizable Fonts Based on Shape Components
  (http://diwww.epfl.ch/w3lsp/publications/typography/pfbosc.pdf)

An example of (b) is the warper code in the autofit module (to be
activated by defining AF_USE_WARPER in aftypes.h) which is highly
dependent on the output resolution since it shifts and scales a glyph
by tiny amounts until a best score is found; this score is based on
how many `segments' (stems and similar features) are near to or
located on grid lines.

Finally, an example of (c) is John Hobby's paper `Generating
Automatically-Tuned Bitmaps from Outlines'
(http://ect.bell-labs.com/who/hobby/89_2-17.pdf).  After analysing the
glyph shapes, you have to solve (sparse) linear equations with a few
hundred to thousand variables, depending on the resolution.

For the ttfautohint project, only (a) is of interest.


    Werner