Closed Dogway closed 1 year ago
About filters I have to update README.md. This is the correct list.
I will take a look at clamp.
I updated README.md with the correct resampling filter names.
libplacebo_Resample(width()2,height()2, filter="ewa_ginseng",radius=6,clamp=0) is still blurrier than libplacebo_Resample(width()2,height()2, filter="ewa_ginseng",radius=3,clamp=0) but not that visibly.
Even libplacebo_Resample(width()2,height()2, filter="ewa_ginseng",radius=3,clamp=1) is visibly blurrier than libplacebo_Resample(width()2,height()2, filter="ewa_ginseng",radius=3,clamp=0).
Usually you should expect blurrier result with clamp=1
.
clamp=1
still can produce negative output values. It clamps only the weights when generating the filter.
You have to manually clamp the output values.
I understand now, only clamps negative lobes(?). I was waiting for this to upload TransformsPack so will update tomorrow. Thanks!
I understand now, only clamps negative lobes(?).
Yes.
Some filter strings don't work on _resample, nor _shader: "box" not working ("nearest" does though) "haasnsoft" also doesn't work
I also had strange results using "clamp"
libplacebo_Resample(width()*2,height()*2, filter="ewa_ginseng",radius=6,clamp=1)
In this case radius=6 is blurrier than radius=3, except when I set clamp to 0. Basically I wanted to clamp negative values so they don't NaN later on float pipelines.