New filters! ⭐⭐⭐#1959
Conversation
|
This PR is not critical for me and has relatively large change for review, so it can wait until 3.4. The most valuable part from my point of view is the filters comparison table in documentation. So if we decide to move this PR to 3.4 I can make another PR with the table only. |
|
Ok, bumping to 3.4. |
| #define IMAGING_TRANSFORM_BILINEAR 2 | ||
| #define IMAGING_TRANSFORM_HAMMING 5 | ||
| #define IMAGING_TRANSFORM_BICUBIC 3 | ||
| #define IMAGING_TRANSFORM_LANCZOS 1 |
There was a problem hiding this comment.
Is there any logical order here? It doesn't appear to be numeric, alphabetical, or order added.
There was a problem hiding this comment.
From less expensive and quality to more. Same as in the comparison table.
There was a problem hiding this comment.
@wiredfool Are you happy with this explanation?
There was a problem hiding this comment.
Yeah, though I expect someone is going to helpfully clean up the order in a PR in a year or two.
|
Looks good to me. |
|
@homm Can you write up basic release notes for this in docs/release/3.4.0.rst? |
|
Yes, sure. But I'd prefere to delay this on the weak before the release, to add all notes at once :) |
|
I'd rather have it in there so that we don't miss it. |
3 participants
This adds two new filters to
resizemethod.BOX is event faster than LINEAR because it has a smaller window (0.5) and produces very sharp images. Applicable for significant downscaling.
HAMMING costs like LINEAR, also produces very sharp images and applicable for downscaling at any range. Doesn't look good for upscaling, though.