help for interlacing problem

[Filipon]

Hello,

I am facing a choppy frame problem when I try to deinterlace this video.
I'm using Tdeint in mode=2, which apparently cleans up the interlacing (no blend traces) but once encoded, the video is jerky in panoramic scenes.
When I try other methods of deinterlacing, I have ghost images appearing.
So, I don't know what to use and I would like to have your opinion.
I am attaching an extract from the VOB.
I use Megui for my encodings.

Thank you in advance for your help.

[Selur]

for the fun of it and since I'm testing some stuff atm. I tried Vapoursynth with:
QTGMC(Bob) to 59.94fps + sRestore to 18fps + Spotless + Crop +mClean + RIFE to 36fps
18 fps probably wasn't low enough, source seemps more like around 12fps.
also tried TIVTC 20 23,976fps + sRestore to 11,98 + Spotless + Crop + mClean + RIFE to 23.976fps.
(The last step using RIFE can't probably be used in MeGui since it's Avisynth only, but other frame interpolation me result in decent outputs too)

Cu Selur

[Filipon]

Thank you for your help Selur.
It may be just an impression, but I find that on your tests, there are always jerky images.
Look for example at 00:09 on the file "18to36" when the image goes from left to right, I have a feeling that it hangs to restart, which seems to me less present in the VOB. There is no way to make it fluid?

In addition, I would like to specify that the videos will then be viewed on a television, so I am looking for a solution to stay in fps that can be easily read by the latter. I'm not sure that the TV will accept 36fps for example.

Thanks again for looking into my problem.

[Selur]

No clue regarding playback compatibility of TVs (haven't owned one for years ).

About the jerkyness. Looking at the content when bobbed, there are tons of dublicates and form time to time forward-backward frames.
Reducing the frame rate does seem to remove most of them, but some are still there, so reducing the frame rate before adding interpolation seems to be the right choice.
So may be using
TFM()
TDecimate(mode=7, rate=10)
or using sRestore with target frame rate 10 is a good idea.
And after that denoising and then adding frame interpolation.

Cu Selur

[Selur]

Playing around with this I could fix a bug in my tool and made a version where I used:
TIVTC to go to 10fps, KillerSpots to remove spots (DeSpot and Spotless are to agressive), Crop, mClean to denoise and RIFE to go for 30fps and resize to get to square pixel.

Cu Selur

[Filipon]

Playing around with this I could fix a bug in my tool and made a version where I used:
TIVTC to go to 10fps, KillerSpots to remove spots (DeSpot and Spotless are to agressive), Crop, mClean to denoise and RIFE to go for 30fps and resize to get to square pixel.

Cu Selur

This rendering is very interesting.
Thanks to you, I have a few avenues to explore to correct the problem I am having. I follow you very grateful!

[jagabo]

It looks like the underlying frame rate is very close to 21000/1001 fps.

###################################
#
# Bob using TFM, doubles frame rate
#
###################################

function TFMBob(clip v, int "pp")
{
    pp = default(pp, 6)

    v.GetParity() ? Interleave(TFM(v, field=1, pp=pp), TFM(v, field=0, pp=pp)) : Interleave(TFM(v, field=0, pp=pp), TFM(v, field=1, pp=pp))
}

###################################


Mpeg2Source("VTS_01_1.d2v", CPU2="ooooxx", Info=3) 
Crop(12,0,-4,-0)
TFMBob(pp=0)

Santiag()
source = last
smooth = SMDegrain(source, tr=5, thSAD=500, refinemotion=true, contrasharp=false, PreFilter=2, mode=0, truemotion=true, plane=4, chroma=false)
TDecimate(smooth, Cycle=60, CycleR=39)  # alternate: TDecimate(smooth, Cycle=60, CycleR=39, clip2=source)

Using the alternate TDecimate will keep the frames before the noise reduction, using the noise reduced clip only to determine how to decimate. Normally one would apply noise reduction after decimating. But here the noise is causing decimation errors.

Note that there is a duplicate combed frame near the start of the video causing a back and forth jerk of the giraffes as they enter the frame. I suspect the original cap had dropped and duplicated frames and that is the cause of most of the difficulties here.

[Selur]

Yup, that looks interessting, but I still would use a lower target frame rate for the decimation (still tons of dublicates).

Note that there is a duplicate combed frame near the start of the video causing a back and forth jerk of the giraffes as they enter the frame. I suspect the original cap had dropped and duplicated frames and that is the cause of most of the difficulties here.

yes, it does seem like there were some drops along the creation of the clip.

[jagabo]

I still would use a lower target frame rate for the decimation (still tons of dublicates).

The remaining duplicates are part of the original film. And I suspect the motion interpolation from 10p to 30p will generate gross distortions with large character movements in other parts of the show.

[Selur]

I agree whatever target frame rate Filipon should use will have to be determined by him.