BFF clips
Asd-g opened this issue · comments
Just to let know that there is issue with BFF clips - Asd-g/AviSynth-BWDIF#5
I skimmed through the code and it seems that VS-BWDIF does take into account the field parameter with respect to the _FieldBased prop. Although I personally have not been able to run tests that match the error you mentioned for the AviSynth version.
Test with this clip.
video = core.ffms2.Source(source=r"BFF 3-2 hard telecine.mkv")
video = core.bwdif.Bwdif(video, field=2)
video.set_output()
Here frame 744 from the current version - https://i.slow.pics/8d0VjRPA.png
Here what the output should be - https://i.slow.pics/4EVnwj6T.png
Zoom 4x to make it visible. Take a look at the ear, the hair edge, the nose...
Seems like you reported the wrong frame, I tried to look out for the closest looking frame, my results lead me to the frame 730 (assuming we are counting frames since 0) - https://i.slow.pics/XReW7zeh.png
import mvsfunc as mvf
video = core.ffms2.Source(source=r"BFF 3-2 hard telecine.mkv")
video = mvf.AssumeBFF(video)
video = core.bwdif.Bwdif(video, field=2).std.Trim(740 ,740 )
video.set_output()
After forcing the field to be based of bottom, it may seem like we're actually dealing the same bug as the avisynth version. Would you mind providing another frame for reference or another sample footage in order to more clearly corroborate this bug?
If the colors look darker, maybe its because ffmpeg took the yuv input and converted it to rgb by force.
From the same clip - for me frame 147 (timestamp 02.452).
Current version - https://i.slow.pics/Dc2MR3EB.png
Correct output - https://i.slow.pics/4rgrRPgF.png
How odd, there seems to be a slight off-sync from my software..
Anyways, the closest frames that I could gather were 143and 145
Here is the whole gallery if you are interested to compare the closest frames that I could detect to 147, actually, the last image of my gallery is, in fact, 147, which, as i said, is off-sync - https://slow.pics/c/j5REAN6j