freetype
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ft] Enabling Subpixel Rendering On freetype-2.7.1


From: Werner LEMBERG
Subject: Re: [ft] Enabling Subpixel Rendering On freetype-2.7.1
Date: Mon, 27 Mar 2017 07:49:05 +0200 (CEST)

>> possible, be aware that some old fonts might `explode' in this
>> mode.
>
> Why would the old fonts "explode"? I thought the "subpixel hinting"
> was contributed from Infinality and it works best with modern fonts
> without good hinting.

This is not correct.  Normal hinting is intended for xppem = yppem.
Subpixel hinting, however, is a special mode to adjust hints in
TrueType fonts so that you get good rendering results for xppem =
3*yppem, i.e., the resolution along the x axis is increased by a
factor of three.

Regarding the `explosion', please read the beginning of Greg's
whitepaper on ClearType.

  https://www.microsoft.com/typography/cleartype/truetypecleartype.aspx

Note that the new ClearType mode from Nikolaus has superseded the
Infinality mode (which was very slow and no longer actively
maintained).

> Is it true that, without "subpixel hinting" FreeType2 will just hint
> fonts to "full pixels" which is exactly the scenario the old fonts
> was developed?

Correct.

> I think the fonts would look sharper that way.

Good B/W hinting is *very* expensive to add to fonts since you have to
take care of the exact pixelation; there are just a handful of
well-paid experts who are doing that, and only a handful of fonts that
have such hints at all – mainly the screen fonts that come with
Windows XP and older.  On the other hand, grayscale hinting can be
done quite well automatically (ttfautohint, VTT).

To get `sharper' rendering you have to increase the sampling frequency
along the x axis; there are various possibilities how to do that.
FreeType's solution is, as mentioned above, to increase the horizontal
resolution, taking care of special situations that would yield to
wrong hinting.  Microsoft's solution is super-sampling (which is the
superior approach IMHO), as explained in the whitepaper.


   Werner

reply via email to

[Prev in Thread] Current Thread [Next in Thread]