[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ft-devel] Splitting up the GSoC project (Re: Freetype-devel Digest,
From: |
Hin-Tak Leung |
Subject: |
Re: [ft-devel] Splitting up the GSoC project (Re: Freetype-devel Digest, Vol 148, Issue 36) |
Date: |
Mon, 29 May 2017 20:36:28 +0000 (UTC) |
Mentioning html and javascript rings some alarms; and so does "mock-up images".
I think you may be thinking that there would be interesting differences for
say, just glyphs A to Z for font X between freetype version P and Q. That's 26
x 2 images, and quite eye-ball-able. put it 26 difference-images on a web page
is easy.
The thing is, most of the 26 pairs are probably going to be identical; with a
few showing small differences (1-pixel edges), and very few with large
differences. One will not learn much about rendering differences from such a
small sample.
Let's think of the ultimate goal: suppose there will be a freetype 2.9, say.
You want to know if there is any important rendering changes against 2.8. So
you throw all ~2000 fonts on a linux system, each with about ~1000 glyphs, at
both, rendering for the 3 modes B/W, 8-bit gray, and LCD horizontal layout, at
two pixel sizes, 15 and 30 (where hinting matters more in one and less in the
other).
2000 x 1000 x 3 x 2 = 12,000,000 , or 12 million pair of images.
A substantial part of that 12 million pairs will be identical; a big part of
the rest will be small number of pixels ar corners, etc; then out of the rest,
most will be 1-pixel edges; and ideally, you only want to actually eye-ball the
remaining. You do not want to eye-ball a handful of pixels differences at the
sharp corners, etc.
I'd suggest:
1. Think of some ways of comparing images which does not involve your eye
balls. You do not want to eye-ball even 1% of 12 million.
2. pick any one font, actually generate 26 images of A to Z, for two versions
of freetype, to get a feeling of what sort of bitmap differences you are likely
to get.
- Re: [ft-devel] Splitting up the GSoC project (Re: Freetype-devel Digest, Vol 148, Issue 36),
Hin-Tak Leung <=