Can't properly DeInterlace or IVTC

[manono]

Either that (try that first) or load them into the script:

LoadPlugin("C:\Path\To\DGDecode.dll")
LoadPlugin("C:\Path\To\TIVTC.dll")
MPEG2Source("C:\Path\To\Movie.d2v")
TFM()
TDecimate(Mode=2)
LanczosResize(720,480)
ConvertToYUY2()

Adjust for your names and paths.

[rbatty11]

jagabo, manono,

...You guys rock! Thank you! Worked perfectly.. When I dl'd TIVTC, I thought i did so into the plugin folder but not so.. duh. Also, I'd have never known to use 24->29.97fps in DGPulldown if you didn't mention it. I would've used the normal 23.976->.. Not sure if that'd have made any difference..??

Just curious, would there have been any noticeable difference if I used TDecimate(Mode=0,Cycle=25,CycleR=1) instead of what I used, TDecimate(Mode=2) ?

Again, thanks for your help.

[jagabo]

Also, I'd have never known to use 24->29.97fps in DGPulldown if you didn't mention it. I would've used the normal 23.976->.. Not sure if that'd have made any difference..??

If your source is 23.976 fps use the 23.976->29.97 pulldown. If your source is 24 use the 24->23.976 pulldown. If you use the wrong one you will get the wrong frame rate. The DVD authoring software may reject it. Or if you succeed in making a DVD it probably won't play properly.

Just curious, would there have been any noticeable difference if I used TDecimate(Mode=0,Cycle=25,CycleR=1) instead of what I used, TDecimate(Mode=2) ?

TDecimate(Mode=0,Cycle=25,CycleR=1) leaves you with 24 fps. TDecimate(Mode=2) leaves you with 23.976 fps.

The former says to remove 1 frame (CycleR=1) out of every 25 (cycle=25). Since you are starting with 25 fps you will be left with 24 fps. The latter says to do whatever is necessary (in this case remove frames) to create 23.976 fps. Both remove duplicate frames whenever possible.

[manono]

Not sure if that'd have made any difference..??

After authoring, the audio would have gotten a little bit out of synch by the end.

[rbatty11]

Well, I used tdecimate(mode=2), which would yield 23.976fps. That means I should've used the 23.976-->30fsp in DGPulldown. No problem as I haven't finished the project yet. I only worked everything with a 2minute clip.

Just have to remember that tdecimate(mode=2) does whatever's need to bring video down to 23.976 vs. TDecimate(Mode=0,Cycle=25,CycleR=1) which would yield 24fps ((CycleR=1) means remove 1 frame).

Thank you!

[igoryonya]

poisondeathray suggested to use TempGaussMC_beta1 if I want a higher quality BOBer.
I downloaded it and all the requirements, that were listed on it's page. Installed all of them in the plugins directory.
I can't get it to work.
VirtualDub comes up with an error:

Avisynth open failure:
Script error: there is no function named "MVAnalyse"
(C:\Program Files\AviSynth 2.5\plugins\TempGaussMC_beta1.avs, line 161)
(H:\utils\avs\restoreBlendedFields.avs, line 7)

---

My script is:

LoadPlugin("C:\Program Files\DGMPGDec\DGDecode.dll")
Import("C:\Program Files\AviSynth 2.5\plugins\TempGaussMC_beta1.avs")
Import("C:\Program Files\AviSynth 2.5\plugins\Srestore.avsi")


MPEG2Source("M:\Media-Edit-Author\video1\VTS_02_1.d2v")
TempGaussMC_beta1()

#SRestore(FRate=24)

----

I've installed mvtools2.dll in the plugins directory.

[manono]

Just have to remember that tdecimate(mode=2) does whatever's need to bring video down to 23.976 vs. TDecimate(Mode=0,Cycle=25,CycleR=1) which would yield 24fps ((CycleR=1) means remove 1 frame).

You can set whatever framerate you want when using Mode=2 by also setting the Rate Parameter. One example might be:

TDecimate(Mode=2,Rate=24)

and you'll get 24fps. Sometimes it helps to read the included docs.

TDecimate(Mode=2)

is just the default setting.

[poisondeathray]

@igoryonya - i think you need the old mvtools, not the v2 branch

[manono]

I've installed mvtools2.dll in the plugins directory.

And V2 of MaskTools also? And RemoveGrain? And Repair?

Yeah, he needs the old MVTools.

[igoryonya]

Yes, poisondeathray, the old one worked.

[poisondeathray]

Be prepared to wait a long time...it is a slow filter

While you are waiting, check out this comparison for bobbers

https://forum.videohelp.com/images/guides/p1934885/stockholma_0-520_q3_yadif_mvbobmod_t...mca4_tdtmm.avi

[igoryonya]

manono, yes, I installed MaskTools 2 and RemoveGrain and Repair.
It took me a few days to gather and read about all the dependencies. Pretty time consuming task, but I am glad that I did it. Looking at the BOB comparison demo that poisondeathray posted, I am really psyched that he told me about TempGaussMC. TempGaussMC is the only BOBer that doesn't jitter.

I am such a nitpicker that I would rather wait for 2 weeks for an HQ video to complete rendering, then be satisfied with an average or LQ video.

I see what you are saying about slow. I get about .2 frames/sec with this bobber with SRestore and deblocker, but I will use it from now on, unless something more advanced or faster with the same quality comes out. I am used to lengthy video rendering wait, because I use NeatVideo to clean up the noise. It gets about .6-.8 frames/sec on this resolution, then I use the Video Enchancer to resize to the correct aspect ratio, which does about 1.5-3 frame/sec on this resolution.

I don't use Lanczos resizer, because the Video Enchancer uses adjacent frames to increase resolution with more detail, not just do sharp resizing as lanczos use, plus, if I would use sharp resizer, I would wait for when S-Spline XL algorithm comes out for video. I was very surprised when I saw the difference between Lanczos and S-Spline XL for the first time. It's as much better then Lanczos as Lanczos is better then bicubic.

[jagabo]

I suspect S-Spline XL will make video look very cartoonish.

[poisondeathray]

I don't understand, are you upscaling for your project? Do you mean S-Spline XL for photos as in photozoom? How does it compare to normal Spline36Resize() ?

While TempGaussMC works well for pure interlaced content, I am unsure how well it will do with your blended frames, because it isn't a true bobber in the sense that the original fields are not preserved - it filters them. Maybe you should test it out with the Srestore() function on a small clip before wasting hours/days on the final project?

[jagabo]

I don't understand, are you upscaling for your project? Do you mean S-Spline XL for photos as in photozoom? How does it compare to normal Spline36Resize() ?

90x64 source:


4x enlarged with Spline36Resize() in (AviSynth):


4x enlarged with S-Spline XL - photo soft mode (Photozoom)


4x enlarged with S-Spline XL - photo detailed mode (Photozoom)


Original higher resolution image (90x64 image was made from this with VirtualDub Lanczos3):

[igoryonya]

I Tested it and I liked it, it's a lot better then the regular bob I used, thanx.
I am just sorry, that I didn't know about it before.
In fact, when I did Yadif and then SRestore, there were still some blended frames on fast motion, but for some reason, there was less blending visible on same frames after the TempGaussMC. I don't know why it is, but I am really happy about it.

I upsize to the correct ratio using Video Enchancer.
I already compared many videos after resizing, using regular, even advanced resizing methods to Video Enchancer method. Nothing compares to it, even the S-Spline XL from PhotoZoom looks less detailed, because Video Enhancer uses adjacent phrames to get more detail nonexistent in just one frame. I mean, regular resizers look sharper, no doubt, but if you look at Video Enhancer results, you see a lot of small detail that was not visible before on that frame.

I don't know what that method is called though.

[poisondeathray]

I'm not a big fan of the splineXL for video, as jagabo said "cartoonish", the edges are destroyed look especially around his left chin, and hairline. I usually don't upscale video, but I find nnedi interpolation + some form of sharpening e.g. LSF + some form of antialiasing usually gives decent results

I just did some testing with Video Enhancer; maybe I'm using it wrong, but the output looks like poor interploation with bad anti-aliasing. Is there some trick to make it work better? I just used the slow HQ method and 2x resize on some test clips

Here is a test from the Clone Wars 1080p Apple Movie Trailer. I chose a CGI cartoon, because the lines and shapes are very clean and you can see jaggies and ringing artifacts very easily. I re-encoded using Spline36 to 240x104 using lagarith (lossless), then used that as the input clip for the tests upscaling 2x to 480x208. This was done on the 2min13sec clip, not stills, so if any temporal filtering was used it should still work, and the artifacts are not from compression because everything was done losslessly

bicubic


spline36 lsfmod


video enhancer


nnedi spline36 lsfmod aaa


bicubic


spline36 lsfmod


video enhancer


nnedi spline36 lsfmod aaa