Provide option to use bilinear filtering for resizing masks/grayscale images #16347
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
The fallback resizing option in
resize_image
is Lanczos, which is a good default in the majority of cases. However, when resizing grayscale images, in particular masks, it can introduce ringing artifacts. A better option in this case is bilinear filtering. You can see the difference in the filters in the provided image (which are outputs of the cropping masks used in img2img during ADetailer processing with mask blurring set to zero).Both nearest and bicubic were considered, however:
To this end, this commit introduces a new option in the img2img settings (disabled by default) that allows the use of bilinear filtering when resizing, if the incoming image is grayscale (PIL's mode 'L'). In all other cases, Lanczos is used as is the case currently.
I've made it a toggle because it does alter the shape of the mask enough to potentially affect current workflows. The effect is also very subtle (as evidenced by the image), but I would argue is more correct.
Screenshots/videos:
Checklist: