Re: [ft-devel] GSoC - Image comparison in test framework

[Alexei Podtelezhnikov]

>>   . A glyph isn't rendered at all due to an internal problem of
>>     handling the glyph's data.
>>
>>   . A stem gets positioned higher or lower; this affects a series
>>     of consecutive horizontal or vertical pixels.  This is the
>>     most frequent – and most important case.
>>
>>   . In B/W mode, a rendered structure has one pixel more or less.
>
> If I'm correct, I believe the MD5 sums should detect these.

Yes, but what's needed is a solution to nicely visualize the
differences.

Indeed, MD5 will not tell how visible are the differences. AFAIK, 64-bit and 32-bit FreeType will produce different results today, but those differences would not be visible. I would appreciate some quantitative measure how visible are the detected differences. Font rendering quality is mostly about contrast, so some integral contrast measure and how it changed would also be nice.