[gs-devel] Re: A CJK expertise wanted

Leonardo leonardo at artifex.com
Tue Feb 21 03:23:46 PST 2006

Dear Mr. Suzuki,

Nice to hear from you.
Your help is really useful.

I would like to know better about the upper horizontal arm of Su.
In the font design space it is not exactly horizontal.
It upgrades from left to right.
Attaching a PDF file which embeds that font.
Use zoom to view it.

The question is whether we need to preserve
the slope in raster when its size is close to the pixel size.

hint2 moves the arm middle to the pixel boundaries,
so that bumps appear at ends, and the slope looks smaller.

hint3 moves the end of the arm to pixel boundary.
Therefore the boundary crosses pixel centers near
the middle of the arm, and in hint3 the left part of the arm
looks lower than the right part of the arm.

So I would say that hint2 diminishes the slope,
and hint3 enlarges the slope.

The question is whether the slope is important for reading.
If it is, we should prefer hint2.
Otherwise we should prefer hint3,
because it looks more smooth.
This questions may look stupid,
but I really need a reasoned answer
because my knowledge of CJK languages is poor.

I guess the slope is not much important with
alphabetic glyphs such as Katakana or Hiragana.
But when we deal with hieroglyphs,
I suspect that it may be opposite.
Therefore I'd appreciate a better explanation
about importance of slopes for CJK hieroglyphs.
Also I'd like to know whether the importance is same for
all CJK languages, or may depend on language.

Thank you.


----- Original Message ----- 
From: <mpsuzuki at hiroshima-u.ac.jp>
To: <leonardo at artifex.com>
Cc: <gs-devel at ghostscript.com>; <tech at artifex.com>; "Raph Levien" 
<raph.levien at gmail.com>
Sent: Tuesday, February 21, 2006 12:45 PM
Subject: Re: [gs-devel] Re: A CJK expertise wanted

>I attached 2 parts that I focused: there are 3 characters,
> "Su", "Te", "Mu" (left to right). In this part, I prefer hint3.
> Check the top horizontal stroke of "Su". There are bumps in hint2,
> and it looks ugly, but no bumps in hint3, so hint3 looks better.
> However, I cannot say hint2 is always better. We can find extra
> dots around both ends of top horizontal strokes of "Su" in hint3.
> If antialiased device, the bumps in hint2 will be quiet.
> The sample is drawn by "Gothic" fonts, it's someting like
> "Helvetica" for Japanese. But, if "Mincho" fonts (like
> "Times-Roman": has serif, the width of strokes are varied),
> there's possibility that hint2 can be better.
> On Mon, 20 Feb 2006 15:20:17 -0800
> "Raph Levien" <raph.levien at gmail.com> wrote:
>>On 2/20/06, Leonardo <leonardo at artifex.com> wrote:
>>> Folks,
>>> Two images attached.
>>> Could anybody please tell me,
>>> in which one the Japanese font looks/reads better ?
>>> Thank you.
>>> The attached ZIP file contains two rasters of same document.
>>> Rasters are created with two working revisions of Ghostscript,
>>> using ppmraw device at 300 dpi.
>>> One can open them with XnView.
>>> The difference is :
>>> hint2.ppm applies a grid fitting to stem poles.
>>> hint3.ppm applies a grid fitting to stem middles.
>>I think hint3 is slightly better, but there are a few glyphs in which
>>I prefer hint2. One example is the hiragana "su", which is the first
>>glyph on the line above the text "FULL STORY>". The "ha" glyph (fifth
>>on that same line) is also better in hint2.
>>Second set, same story. Most glyphs are better in hint3, but in hint2
>>I like the kanji after the text "SEK 2003" better because the stem
>>width is more consistent.
>>Hope that's helpful.
>>gs-devel mailing list
>>gs-devel at ghostscript.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: japan--.pdf
Type: application/octet-stream
Size: 154835 bytes
Desc: not available
Url : http://ghostscript.com/pipermail/gs-devel/attachments/20060221/6c51ab0c/japan---0002.obj

More information about the gs-devel mailing list