@manono/all Clerks IVTC with Decomb

[Mechannibal]

Hi,

I am having problems with IVTC in the movie Clerks.

Manono had mentioned in this thread, which related to IVTC issues, "Damn Kevin Smith. I had similar problems with Clerks. When's he going to have enough money to make his movies properly?"
http://forum.doom9.org/showthread.php?s=&threadid=18823&highlight=clerks

I am curious what Manono used to get a proper IVTC with Clerks?

I had found Forced Film would sometimes have combed frames in the video sequences, so I tried Decomb 4.01 with:

telecide(post=true)
decimate(cycle=5)

and

telecide(post=false)
decimate(cycle=5)

and

telecide(post=false,guide=1)
decimate(cycle=5)

All of which gave good results during the video sequences, but during the title sections with black background and white writing, I noticed a few combed frames and the text would bob up and down for a second.

In testing I looked specifically at the part that says 'View Askew Productions Presents' in white lettering in the beginning.

Any advice from anyone is welcome.

Thanks,

Mechannibal

[manono]

Hi-

Man, that was an old thread. I don't even remember what settings I used. I would do the movie differently if I were doing it today (I'd make my crop settings all even numbers, for one thing). I was trying to put it on 1 CD, and it compresses horribly. And I wound up smoothing it too much. But back then we didn't have Vlad59's wonderful Convolution3D to work with either.

I have a bit of that shimmer in the text also, and one answer I'm going to purpose probably won't help that, and might make it worse. But, to help get rid of the interlacing, you might strengthen the deinterlacing settings to something like:

Telecide(guide=1,threshold=10,dthreshold=5)
Decimate(cycle=5)

And if that doesn't work, then you might consider trying IVTC2.2 or IT to see if one of them can handle it any better.

[Kb_cruncher]

I have only had to IVTC once using a different method other than DVD2AVI's force film.After trying every deInterlace method i could find,i found that GreedyHMA Gave me perfect results.Gordian Knot comes with it by default.It uses motion compensated deInterlace and IVTC.


GreedyHMA(1,0,4,0,0,0,0,0)#was the parameter i used.The first parameter(1)calls top/bottom field first.The third parameter is the deInterlace/pulldown method.4=Auto pulldown with decimation.

NOTE:I had some strange results using MPEG2DEC3 while trying to do IVTC.

hope it works for you as it did me.

[Dagon]

How funny, I'm trying to encode Clerks this weekend too. I'll tell you how it comes out. It's got so much film grain! And it IS the worst compressing movie I've ever done!....:scared: Going to try Convolution3D.

[ErMaC]

Don't forget to encode it to XviD and use it's "Encode Grayscale" flag. :)

[Mechannibal]

Here's the followup:

I tried manono's suggestion:
Telecide(guide=1,threshold=10,dthreshold=5)
Decimate(cycle=5)

And it ended up looking worse on some parts.
I had tried IVTC2.2 and IT with the help of someone who was familiar with those, and they also produced worse results. I also tried GreedyHMA(1,0,4,0,0,0,0,0) and it produced worse output as well.

So I settled with:

telecide(post=true)
decimate(cycle=5)

which still had shimmering text with a few frames of text that looked interlaced, but was the best of the bunch.

When using the chosen setting I noticed some other irregularities. In Clerks, when the titles of the chapters would appear with white lettering and black background the frames directly before and after the the titles would be shifted either up or down from the previous or subsequent frames. The part shifted up and down was the actual movie, not titles.

So to illustrate:
frames leading up to the title would be fine except for the last frame before the scene change; it would be shifted.
also, the same happened right after the titles, the 1st frame after the scene change would be shifted up or down one from the subsequent frames.

This created a slight shimmer effect before and after the titles that is barely noticeable, in addition to the occaisional shimmer of the titles.

I had not used any filtering on my encode. I used Xvid 04102002 stable Koepi's compile with H263 for both passes, and I enabled the greyscale option. I was going for 1 cd with a target filesize of 627939 kb. The resulting file was 627980 kb.
Other settings were defaults except for lower quants for credits.


My AVS:

LoadPlugin("D:\DVDRIP~2\GORDIA~1\mpeg2dec.dll")
LoadPlugin("D:\DVDRIP~2\GORDIA~1\decomb.dll")

mpeg2source("D:\DVD Rip\Clerks\Clerks -encode\Clerks yuv176 nofFilm 29.97fps.d2v")

Telecide()
Decimate(cycle=5)

crop(3,66,715,342)
BicubicResize(608,320,0.333,0.333)


and here is my analyzed debugview file:

XviD Analyzer v0.12 by MoonWalker & MarcFD
e-mails : s_ilias@gmx.net & marc.fd@libertysurf.fr


Quantizers Analisis
---------------------

Quantizers Used For Movie :
------------------------------
Quant 2 Used : 1360 Times, Percentage Used : 1.06%
Quant 3 Used : 507 Times, Percentage Used : 0.40%
Quant 4 Used : 1543 Times, Percentage Used : 1.21%
Quant 5 Used : 30189 Times, Percentage Used : 23.62%
Quant 6 Used : 77291 Times, Percentage Used : 60.46%
Quant 7 Used : 13584 Times, Percentage Used : 10.63%
Quant 8 Used : 1713 Times, Percentage Used : 1.34%
Quant 9 Used : 909 Times, Percentage Used : 0.71%
Quant 10 Used : 39 Times, Percentage Used : 0.03%
Quant 11 Used : 667 Times, Percentage Used : 0.52%
Quant 12 Used : 8 Times, Percentage Used : 0.01%
Quant 13 Used : 2 Times, Percentage Used : 0.00%
Quant 14 Used : 1 Times, Percentage Used : 0.00%
Quant 16 Used : 18 Times, Percentage Used : 0.01%

Average Quantizer Used for Movie : 5.869

Quantizers Used For Credits :
--------------------------------
Quant 10 Used : 14 Times.
Quant 18 Used : 4140 Times.

H.263 Quantization Type Used 131985 timed, Percentage Used : 100.00%

Quantizers prevented from rising too steeply 1314 times


Intra-Frame (Key-Frame) Quantizers
------------------------------------

Movie
-------
Quant 2 Used : 24 Times, Percentage Used : 3.39%
Quant 3 Used : 1 Times, Percentage Used : 0.14%
Quant 11 Used : 649 Times, Percentage Used : 91.54%
Quant 12 Used : 3 Times, Percentage Used : 0.42%
Quant 13 Used : 2 Times, Percentage Used : 0.28%
Quant 14 Used : 1 Times, Percentage Used : 0.14%
Quant 16 Used : 15 Times, Percentage Used : 2.12%

Credits
---------
Quant 10 Used : 14 Times, Percentage Used : 1.97%

Inter-Frame (P-Frame) Quantizers
------------------------------------

Movie
-------
Quant 2 Used : 1336 Times, Percentage Used : 1.02%
Quant 3 Used : 506 Times, Percentage Used : 0.39%
Quant 4 Used : 1543 Times, Percentage Used : 1.18%
Quant 5 Used : 30189 Times, Percentage Used : 23.00%
Quant 6 Used : 77291 Times, Percentage Used : 58.88%
Quant 7 Used : 13584 Times, Percentage Used : 10.35%
Quant 8 Used : 1713 Times, Percentage Used : 1.30%
Quant 9 Used : 909 Times, Percentage Used : 0.69%
Quant 10 Used : 39 Times, Percentage Used : 0.03%
Quant 11 Used : 18 Times, Percentage Used : 0.01%
Quant 12 Used : 5 Times, Percentage Used : 0.00%
Quant 16 Used : 3 Times, Percentage Used : 0.00%

Credits
---------
Quant 18 Used : 4140 Times, Percentage Used : 3.15%



Frame Analisis
----------------

Number Of Intra-Frames (Key-Frames) : 709
Number Of Inter-Frames (P-Frames) : 131276
Total Number Of Frames : 131984

0.54% of the Movie is Intra-Frames (Key-Frames)
99.46% of the Movie is Inter-Frames (P-Frames)



Size Analysis
----------------

1-Pass Size : 3657210018 Bytes or 3571494 KBytes or 3487 MBytes
Scaled Size : 639822744 Bytes or 624826 KBytes or 610 MBytes
Actual Size : 639811055 Bytes or 624815 KBytes or 610 MBytes

Usefull Statistics
------------------
Compressibility : -100.32%
Relative Quality of XviD avi : 34.08%
Absolute Quality of XviD avi : 88.39%


With an average quant of 5.869, the movie compressed very badly. With post processing on the x and y luma planes it looks ok, nothing spectacular, but no big problems either.

I hope that other encoders can use my example to save time with this movie. I will probably reencode this movie with some sort of filter to improve compressibility.

Thanks for your suggestions.

Mechannibal