Using TBC and video still tearing?

[jriker1]

I was curious sanlyn. In the original AVISynth scripts (or the 4th one) you created for HHGTTG, you have:

AssumeTFF().TFM().TDecimate()

I was always doing BFF with the DV input assume the analog ATI is TFF? I'm going back to some of the bad content, especially Hitchhikers one as the DVD version of this video is missing some good parts so decided to go back to trying to cleanup this VHS version. Started to step thru it and saw that so was curious.

Also in your two step process the first step you say to save as Lagarith YV12. How about the second one. I try to avoid YV12 as much as possible leaving as YUY2 but understand with some of the scripts you use this may be necessary.

Thanks.

JR

[jriker1]

Forgot one part. On the first script again, is it assumed the footage is progressive?

Import("D:\Avisynth 2.5\plugins\ContrastMask.avs")
DirectShowSource("E:\forum\jriker1\Cap1\HHGTTG.avi ")
Trim(0,122)TComb()
CheckMate(24)
Santiag()
AssumeTFF().TFM().TDecimate()
ConvertToYV12(interlaced=false)

So above, last line of what I copied in here, you assume the footage is not interlaced. Curious why.

Is the assumption the footage was originally progressive and TFM is returning it back to that state? From what I can see TFM has it's own deinterlacer for footage it can't pull a progressive frame out of. Is it better to use one of the better deinterlacers like QTGMC as part of this process? Looks like TFM supports external deinterlacers.

Thanks.

JR

[sanlyn]

Quote:

Originally Posted by jriker1

I was curious sanlyn. In the original AVISynth scripts (or the 4th one) you created for HHGTTG, you have:

AssumeTFF().TFM().TDecimate()

I was always doing BFF with the DV input assume the analog ATI is TFF?

DV is always BFF -- supposed to be, anyway. Analog tape is almost always TFF, and so is DVD/BluRay/AVCHD. But always test first. Everyone doesn't follow the standard. Analog tape in TFF is captured by DV as BFF. Some people say that the process "makes no difference". I say it screws up everything, including something of the field phasing in the original interlacing. VHS-->DV always looks weird to me, but many worship the results as manna from DV heaven. I disagree, but maybe it's just me.

Quote:

Originally Posted by jriker1

Also in your two step process the first step you say to save as Lagarith YV12. How about the second one.

IF the output of step1 is saved as YV12, it will still be YV12 when used as input for step2. Most of the filters in step 2 require YV12. What most people do with YUY2 input is process as much as possible in the original colorspace, then properly convert to the next required colorspace rather than go back and forth -- if at all possible, depending on the sequence of filters required. Proper conversion demands that field/frame structure is observed -- something many NLE's don't do. Because Avisynth is programmed to make those conversions with as little damage as possible, I almost always use Avisynth to do it.

Quote:

Originally Posted by jriker1

Forgot one part. On the first script again, is it assumed the footage is progressive?

Import("D:\Avisynth 2.5\plugins\ContrastMask.avs")
DirectShowSource("E:\forum\jriker1\Cap1\HHGTTG.avi ")
Trim(0,122)TComb()
CheckMate(24)
Santiag()
AssumeTFF().TFM().TDecimate()
ConvertToYV12(interlaced=false)

So above, last line of what I copied in here, you assume the footage is not interlaced. Curious why.

The original motion picture film is always progressive. What people do to it later is a matter of infinite variety. After hard telecine is undone by IVTC, the video is progressive. What remains are some field-blended frames, which are effectively "progressive" in nature and usually can't be unblended.
The thing to remember about the above code is that all of the filters in Step1 can work in both YUY2 and YV12.

Quote:

Originally Posted by jriker1

Is it better to use one of the better deinterlacers like QTGMC as part of this process? Looks like TFM supports external deinterlacers.

TIVTC looks for telecined frames and telecine patterns. If you deinterlace first, TFM won't interpret the telecine patterns correctly because the combination of deinterlaced-progressive and deinterlaced-telecined frames -- i.e, most of the frames won't appear to be telecined. Yes, you can use something like QTGMC first, but you have to reduce the number of duplicate fields by using SelectEven() or SelectOdd(), whichever gives the most consistent results, then saving that new clip and feeding it to TFM/TDecimate. This is often done for videos that have really odd duplicate/blended/telecined frames, which you'll see in some anime shot originally at 15fps, and often in conjunction with a few deblenders (moist of which are not effective on many screwed up videos) or sRestore. You'll also see it in many of the godawful corrupt videos from India.

[jriker1]

Quote:

Originally Posted by sanlyn

TIVTC looks for telecined frames and telecine patterns. If you deinterlace first, TFM won't interpret the telecine patterns correctly because the combination of deinterlaced-progressive and deinterlaced-telecined frames -- i.e, most of the frames won't appear to be telecined. Yes, you can use something like QTGMC first, but you have to reduce the number of duplicate fields by using SelectEven() or SelectOdd(), whichever gives the most consistent results, then saving that new clip and feeding it to TFM/TDecimate. This is often done for videos that have really odd duplicate/blended/telecined frames, which you'll see in some anime shot originally at 15fps, and often in conjunction with a few deblenders (moist of which are not effective on many screwed up videos) or sRestore. You'll also see it in many of the godawful corrupt videos from India.

Thanks sanlyn. I may have not written things clearly. What I read is TFM deinterlaces when it has to with some internal deinterlacer. So in the event TFM decides to deinterlace in your original provided code, it uses whatever it uses. Don't recall what I read on when it chooses to deinterlace. You can tell TFM to use some other deinterlacer if it feels deinterlacing is required with something like: AssumeTFF().TFM(Slow=2, clip2=QTGMC(preset="very slow")).TDecimate() with the clip2 parameter. Plus add SelectEven() or SelectOdd() as you referenced to the QTGMC call as well. Not sure if having the clip at 59.97 is bad or will just cause problems with the remaining steps of the scripts. I'm deinterlacing to 59.97fps with my other clips that are more just standard interlaced.

- Import to AVI Lagarith
- AVISynth with just:

import("QTGMC-3.32.avsi")
AVISource("filename", pixel_type="YUY2")
AssumeTTF()
QTGMC(preset="very Slow")

- Then move the resulting file into Premiere Pro and use NeatVideo to clean it up and chop off the edges that have flaws. Save to H264.

JR

[sanlyn]

Quote:

Originally Posted by jriker1

Thanks sanlyn. I may have not written things clearly. What I read is TFM deinterlaces when it has to with some internal deinterlacer. So in the event TFM decides to deinterlace in your original provided code, it uses whatever it uses. Don't recall what I read on when it chooses to deinterlace. You can tell TFM to use some other deinterlacer if it feels deinterlacing is required with something like: AssumeTFF().TFM(Slow=2, clip2=QTGMC(preset="very slow")).TDecimate() with the clip2 parameter. Plus add SelectEven() or SelectOdd() as you referenced to the QTGMC call as well. Not sure if having the clip at 59.97 is bad or will just cause problems with the remaining steps of the scripts. I'm deinterlacing to 59.97fps with my other clips that are more just standard interlaced.

- Import to AVI Lagarith
- AVISynth with just:

import("QTGMC-3.32.avsi")
AVISource("filename", pixel_type="YUY2")
AssumeTTF()
QTGMC(preset="very Slow")

- Then move the resulting file into Premiere Pro and use NeatVideo to clean it up and chop off the edges that have flaws. Save to H264.

JR

If an avsi plugin is in your plugins folder, you don't have to "import" it. avsi's load automatically.

QTGMC requires YV12.
"Very slow" is very strong filtering ans usually over smooths. Add NeatVideo and your video will look like a posterized cartoon.

Deinterlacing telecined/blended video results in duplicates of progressive frames, duplicates of telecined frames, and duplicates of blended frames. Many players will choke on that at 59.96fps.

Works only with purely interlaced video.

[jriker1]

Thanks for the info. For reference, QTGMC is YUY2 or YV12.

http://avisynth.nl/index.php/QTGMC

See requirements on the right.

I was originally concerned about "very slow" and the developer said it was light cleaning but will probably back off to something less intense. Wanted the best quality deinterlacing but saw also that it adds some level of cleaning automatically. Not sure why that choice was made by the creator.

JR

[sanlyn]

Quote:

Originally Posted by jriker1

Thanks for the info. For reference, QTGMC is YUY2 or YV12.

http://avisynth.nl/index.php/QTGMC

See requirements on the right.

Only with Avisynth 2.5x and mt_masktools v2.0.45.0 (September 2010). It's the only version of mt_maskools2 that works with YUY2. Internally it does a planar conversion and takes forever.

Quote:

Originally Posted by jriker1

I was originally concerned about "very slow" and the developer said it was light cleaning but will probably back off to something less intense. Wanted the best quality deinterlacing but saw also that it adds some level of cleaning automatically. Not sure why that choice was made by the creator.

It depends on the video's condition.

[jriker1]

So I am still here.

I have all the tapes converted to lossless digital format and paused for a while. I have now loaded the first one into Premiere Pro and have used an adjustment layer to clean up the coloring thruout the various pieces. Plus using NeatVideo for cleanup. I am going to export TFF interlaced and then load into QTGMC thru AVISynth to deinterlace into it's final format. Question I have right now, what should it's final format be? I notice a lot of codec come and stop being supported and as MS updates their OS versions, some won't work anymore. Any recommendations for what codec and container I should be using to archive these materials? Not concerned with lossless as long as I can maintain decent quality being VHS originals of course.

Thanks.

JR