Hi Jeremy,
I looked through the comments in the bug and your wiki. If
Image.getScaledInstance() is not recommended, we may want to do
something about it: as the bug report suggests, either deprecate the
method and provide guidelines for the recommended way or fix the
implementation.
Yet I don't have extensive experience in Java 2D.
You can add this a the comment in JBS and ask for Phil's opinion.
Regards,
Alexey
On 23/06/2023 07:06, Jeremy Wood wrote:
The performance of Image.getScaledInstance() has been known to be a
problem for a long time. (This ticket
<https://bugs.openjdk.org/browse/JDK-6196792> discusses it well. It
has a complicated history.)
I have developed code
<https://github.com/mickleness/pumpernickel/wiki/Scaling> that is much
more performant. It takes less than 10% of the time the
AreaAveragingScaleFilter does. There’s nothing especially clever about
it; it just iterates over the pixels more efficiently. Like
AreaAveragingScaleFilter: it doesn’t suffer degraded appearance
(pixelation) at less than 50% scale.
Is there any interest in me working on a PR to overhaul
AreaAveragingScaleFilter? (That is: is there interest in
reviewing/sponsoring such a PR?) It won’t be a trivial task, but the
results I have now indicate the performance gains could be significant.
(Or in my wildest of pipe dreams: I’d love it if
Graphics2D.drawImage(..) could scale images past 50% without looking
pixelated… but that could be an even more difficult discussion about
RenderingHints and long-established interpolation norms. So I assume
that’s a non-starter.)
Regards,
- Jeremy
