[luatex] Get area of a glyph?

[Peter Rolf]

Am 16.09.2012 17:56, schrieb Arno Trautmann:
> Paul Isambert wrote:
>> Arno Trautmann <Arno.Trautmann at gmx.de> a écrit:
>>> Hi all,
>>>
>>> again a rather unusual question: I'd like to get the black area of a
>>> glyph. Why? To estimate it's greyness: For a “–” e.g. I want to get
>>> something like “5%” (of the whole box surface), while a “m” would give
>>> some “30%”, and a black box of course would result in 100%. Of course I
>>> would also be happy with any absolute value like “0.1 square inches” or
>>> whatever. So: Is there any chance to read such a value from inside
>>> LuaTex? Or is this information not available?
>>
>> LuaTeX doesn't even read the curves used to draw a glyph, so no way you
>> can get that information, as far as I can tell; you'll have to read
>> directly from the font file
> 
> I heard the rumor that LuaTeX can read several things from font files ;)
> So this sounds as if I had to learn how to read a font file and to
> extract the curves in LuaTeX. Will be my task for the next few months …
> 
>> (under the assumption that knowing the shape
>> of the glyph gives you its area, which I don't know, not being a
>> mathematician;
> 
> Being a physicist, I guess it should give the area. But it won't be easy …
> 
>> perhaps by hit or miss?).
>>
>> Perhaps approximations exist, but I've never heard of them.
> 
> Maybe someone else?
>
Your 'per glyph' approach ignores kerning and composed characters.
I doubt that the results would be meaningful, if you ignore what
typesetting is about. I also doubt, that the wanted information is
available in a font. To summarize it: this approach is based on insecure
data and also hard to implement (in any case no fun).

Sounds like a dead end to me. The best way I can think of, is to create
a standard pdf document and then use ImageMagick or GraphicMagick to
convert that pdf into a b&w only png graphic (600ppi or higher). The
conversion itself is done by ghostscript, but I guess you can also use
the graphic tools to count the black points. If that is not possible,
export as a RAW graphic and count the black pixels per algorithm.

No font information or tricky algorithms needed. The other advantages of
this approach are:

- it's universal (create the same PDF with Word, Q-Express, whatever and
compare the gray scale values)
- low error (if resolution is high enough), as you get what you see
- fast

The hardest part is probably to figure out what standards exists to
measure the gray scale of a given document.

Hope that helps,


Peter