DFTTest does not respect user-specified sigma
shssoichiro opened this issue · 2 comments
shssoichiro commented
For example, when calling vsdenoise.DFTTest(clip).denoise(sigma=0.2) (or just to be extra careful, vsdenoise.DFTTest(clip, sigma=0.2).denoise(sigma=0.2) also produces identical results), the video clip is filtered as if the default sigma of 8.0 is used. In most sane cases, this results in denoising that is stronger than expected.
Image 1: dfttest2.DFTTest(clip, sigma=0.2, backend=dfttest2.Backend.CPU)
Image 2: vsdenoise.DFTTest(clip).denoise(sigma=0.2)
The expectation is that these should be identical.
Setsugennoao commented
Are you on git latest?
shssoichiro commented
Yes, double checked that Windows doesn't have some double-installed python or something as well.