hydra3333
27th March 2017, 14:26
Hello, I am seeking your advice on which deinterlacer to use, rather than QTGMC as I use now and which is VERY slow, for the following situation
(pending arrival and evaluation of a new gpu accelerated deinterlacer "DGFieldDeinterlace")
- input either (a) PAL 1080i TFF possibly with resize to PAL 576i TFF or not, or (b) is 576i TFF already, delivered from source filter "DGSource" ... broadcaster-mangled sources
- fast sports action on a large grassed field with lots of panning and zoom-ins, eg per clips like "SAMPLE2.mpg" in this folder https://drive.google.com/open?id=0B5RV2aJ2vdhSUGFvVm1tYnJhejg
- after deblocking and denoising and sharpening etc, if and as required, then re-interlace the final result so as to retain "motion fluidity"
- yields the most reasonable and "sharpest" result
- faster than QTGMC :)
1. DGDecodeNV (uses the GPU decoder and deinterlacer on some Nvidia video cards)
video = core.avs.DGSource(r'test.dgi',deinterlace=2) # deinterlace=2 means double rate deinterlacing, beware extra frame
# If using double-framerate NV, fix the double-framerate bug in NV per http://forum.doom9.org/showthread.php?p=1391556#post1391556
video = core.std.Trim(video,first=1) # *** fix a double-framerate bug in NV per http://forum.doom9.org/showthread.php?p=1391556#post1391556 ***
Pros ?
- very fast
- inbuit in DGDecodeNV
- options to resize at the same time
Cons ?
- was the nvidia deinterlacer said to be a tad soft ? not sure.
2. https://github.com/HomeOfVapourSynthEvolution/VapourSynth-Yadifmod
https://forum.doom9.org/showthread.php?t=171028
yadifmod.Yadifmod(clip clip, clip edeint, int order[, int field=-1, int mode=0])
core.yadifmod.Yadifmod(clip, edeint=nnedi3.nnedi3(clip,field=3), order=1, mode=1, opt=0)
Pros ?
- said to be good, I suppose it must be "nicer" than nnedi3 alone given it uses nnedi3 for predictions
Cons ?
- not so fast, given it also uses nnedi3 for predictions ?
3. https://github.com/dubhater/vapoursynth-nnedi3
https://forum.doom9.org/showthread.php?t=166434
nnedi3.nnedi3(clip clip, int field[, bint dh=False, int[] planes=[0, 1, 2], int nsize=6, int nns=1, int qual=1, int etype=0, int pscrn=2, bint opt=True, bint int16_prescreener=True, bint int16_predictor=True, int exp=0, bint show_mask=False])
core.nnedi3.nnedi3(clip, field=3, exp=2)
Pros ?
- don't know
Cons ?
- don't know
4. https://github.com/HomeOfVapourSynthEvolution/VapourSynth-EEDI2/
https://forum.doom9.org/showthread.php?t=171136
"not really for deinterlacing nowadays"
Pros ?
- don't know
Cons ?
- own documentation says "not really for deinterlacing nowadays"
5. http://www.vapoursynth.com/doc/plugins/eedi3.html
(core plugin in vapoursynth)
eedi3.eedi3(clip clip, int field[, bint dh=0, int[] planes=[0, 1, 2], float alpha=0.2, float beta=0.25, float gamma=20, int nrad=2, int mdis=20, bint hp=0, bint ucubic=1, bint cost3=1, int vcheck=2, float vthresh0=32, float vthresh1=64, float vthresh2=4, clip sclip])
core.eedi3.eedi3(clip, field=3)
Pros ?
- don't know
Cons ?
- don't know
6. Another deinterlacer ?
If there's a link with advice somewhere already, please let me know.
Thanks.
(pending arrival and evaluation of a new gpu accelerated deinterlacer "DGFieldDeinterlace")
- input either (a) PAL 1080i TFF possibly with resize to PAL 576i TFF or not, or (b) is 576i TFF already, delivered from source filter "DGSource" ... broadcaster-mangled sources
- fast sports action on a large grassed field with lots of panning and zoom-ins, eg per clips like "SAMPLE2.mpg" in this folder https://drive.google.com/open?id=0B5RV2aJ2vdhSUGFvVm1tYnJhejg
- after deblocking and denoising and sharpening etc, if and as required, then re-interlace the final result so as to retain "motion fluidity"
- yields the most reasonable and "sharpest" result
- faster than QTGMC :)
1. DGDecodeNV (uses the GPU decoder and deinterlacer on some Nvidia video cards)
video = core.avs.DGSource(r'test.dgi',deinterlace=2) # deinterlace=2 means double rate deinterlacing, beware extra frame
# If using double-framerate NV, fix the double-framerate bug in NV per http://forum.doom9.org/showthread.php?p=1391556#post1391556
video = core.std.Trim(video,first=1) # *** fix a double-framerate bug in NV per http://forum.doom9.org/showthread.php?p=1391556#post1391556 ***
Pros ?
- very fast
- inbuit in DGDecodeNV
- options to resize at the same time
Cons ?
- was the nvidia deinterlacer said to be a tad soft ? not sure.
2. https://github.com/HomeOfVapourSynthEvolution/VapourSynth-Yadifmod
https://forum.doom9.org/showthread.php?t=171028
yadifmod.Yadifmod(clip clip, clip edeint, int order[, int field=-1, int mode=0])
core.yadifmod.Yadifmod(clip, edeint=nnedi3.nnedi3(clip,field=3), order=1, mode=1, opt=0)
Pros ?
- said to be good, I suppose it must be "nicer" than nnedi3 alone given it uses nnedi3 for predictions
Cons ?
- not so fast, given it also uses nnedi3 for predictions ?
3. https://github.com/dubhater/vapoursynth-nnedi3
https://forum.doom9.org/showthread.php?t=166434
nnedi3.nnedi3(clip clip, int field[, bint dh=False, int[] planes=[0, 1, 2], int nsize=6, int nns=1, int qual=1, int etype=0, int pscrn=2, bint opt=True, bint int16_prescreener=True, bint int16_predictor=True, int exp=0, bint show_mask=False])
core.nnedi3.nnedi3(clip, field=3, exp=2)
Pros ?
- don't know
Cons ?
- don't know
4. https://github.com/HomeOfVapourSynthEvolution/VapourSynth-EEDI2/
https://forum.doom9.org/showthread.php?t=171136
"not really for deinterlacing nowadays"
Pros ?
- don't know
Cons ?
- own documentation says "not really for deinterlacing nowadays"
5. http://www.vapoursynth.com/doc/plugins/eedi3.html
(core plugin in vapoursynth)
eedi3.eedi3(clip clip, int field[, bint dh=0, int[] planes=[0, 1, 2], float alpha=0.2, float beta=0.25, float gamma=20, int nrad=2, int mdis=20, bint hp=0, bint ucubic=1, bint cost3=1, int vcheck=2, float vthresh0=32, float vthresh1=64, float vthresh2=4, clip sclip])
core.eedi3.eedi3(clip, field=3)
Pros ?
- don't know
Cons ?
- don't know
6. Another deinterlacer ?
If there's a link with advice somewhere already, please let me know.
Thanks.