Re: [Devel] Questions about rotated characters at small point sizes.

[Keith Packard]

Around 17 o'clock on Apr 27, "Greg Inman" wrote:

> After calling FT_Get_Glyph() to get the glyph image I am calling
> FT_Glyph_Transform() with a rotation matrix if the angle is not 0.  I am
> not giving this function any 2d translation vector.  Then I am calling
> FT_Glyph_To_Bitmap() to render a monochrome bitmap and giving it a
> translation vector.  This translation vector is composed only of the
> fractional parts of the location vector

My understanding is that the coordinates modified by the hints after
applying the matrix set by FT_Set_Transform and that beyond that, the
coordinates are sent straight through the matrix set by
FT_Outline_Transform.

I suspect you're getting the transformation applied after the hints, which 
will probably generate sub-optimal results.

In any case, the Type-1 rasterizer in FreeType won't match the quality of 
the Adobe rasterizer from Display PostScript, especially when you rotate 
the glyphs.  One way to reduce the visible noise would be to use 
anti-aliasing, but that can be a pain without the Render extension.

Sun will eventually support Render in their server as they're moving to 
Gtk+ based applications which will soon essentially require it.

address@hidden   XFree86 Core Team              SuSE, Inc.