Need help from people smarter than me to IVTC a proprietary pulldown

[ExperimentalAnimations]

I have a panasonic GH1 video DSLR which shoots 23.976 but puts its in a 60i wrapper. There are various paid solutions for removing panasonic's 3:2 pulldown, but from my experience they have been unstable, subpar, or have produced very large files. Really the issue is that all of them have combing artifacts in them, which means they're all deinterlacing the footage instead of doing a proper IVTC which shouldnt have any signs of interlace.

Here's a 10sec 14MB .mts sample (http://filesurf.ru/206573). Please take a look and see if there's any way to retrieve the pristine 24p source from it.

[Didée]

Glancing at the naked source, it seems like a basic TFM().TDecimate() should do the job?

[Hagbard23]

Smartdecimate should do the job easily either...(but first choice would be TFM+TDecimate)...

[ExperimentalAnimations]

Nope, still combing artifacts. And some weird luma distortions too.

[poisondeathray]

maybe you have other problems , like system or decoder issues ?

What source filter ? decoder are you using ?



tfm().tdecimate() works fine on that clip for me

I used to have a GH1, and it worked fine for all 24p/60i footage from it

Note: there is a GH1 firmware hack that allows you to shoot 24p native now, and increased bitrates .

[ExperimentalAnimations]

Huh, you're right, it works fine on my most recent clips, but not my older ones. Odd. I've tried the hack, 24p is super unstable, thats why im trying to find a pulldown method.

So how do i batch a folder through this script? Im using virtualdub but i believe it always works in rgb, so if i set it to yuy2 it would do yv12->rgb->yuy2? If so i'd like to find another way. Basically im looking to remove pulldown and convert them to yuy2 lagarith or huffyyuv.

[poisondeathray]

vdub only does the rgb conversion if you use full processing mode, or some vdub filters . If you set to fast recompress it will skip the rgb conversion

you can batch with various avs batch script creators , and vdub in batch mode, or use ffmpeg batch file for huffyuv

if you're not comfortable with commandline and would like some gui tools you can look at this post , search later in the thread, I think some utilites are updated
http://www.hv20.com/showthread.php?25044-Batch-Intermediate-Creation-Utilities

[ExperimentalAnimations]

Ouch, statement redacted, this doesnt work at all, it actually causes the video to glitch and stutter. It only seems to work on some videos and not others. This doesnt make any sense. Ill upload a clip once i can be certain its not a configuration issue.

[ExperimentalAnimations]

My 60i is now having pulldown and deinterlacing done to it in vdub, without any filters applied to it! I dont know what else to say, WHAT THE HELL.

[Guest]

Did you by chance resize it to half size in VirtualDub, using either the resize filter or by changing the display size?

If not, is it the same clip in the first post? That one has hard pulldown so there's no way it could be corrected in VirtualDub without filtering.

And how are you opening the MTS in VirtualDub? Are you using Avisynth? If so, what is the script?

You're giving us such a moving target here that it is hard to know exactly what you are doing. Please give us the video clip and all other details needed.

[ExperimentalAnimations]

Info() is telling me that vdub reads it as progressive even though it is most definitely interlaced as i just recorded it.

Heres a new test file (20MB) (http://filesurf.ru/206744). I tried the Tdecimate filter and it just gives me combing artifacts, though i wouldnt be able to tell anyways as vdub seems to do what it wants.

edit: Does this clip have hardpulldown? Im not sure what the difference is, but its just 24p in a 60i wrapper. I didnt resize or do any processing, display size is at 100%.

[Guest]

Please say what script you use. If no script, then how do you open MTS in VirtualDub? Did you install a plugin? Are you resizing as I asked? Please answer all my questions.

Downloading your clip.

EDIT: That server is dog slow. Use mediafire.com.

[ExperimentalAnimations]

Script is

Directshowsource("C:\....\myfile.mts")

Ive also checked my directshow filters (ffdshow) to see if its doing processing and everything is unchecked. Graphedit is showing my decompressor as just "avi decompressor" with little "xforms in/outs" on it, which goes out to VMR video renderer. Unless you see any problems with that i dont know what else to do.

edit: mediafire just gives me a "failed upload, IO error".

[poisondeathray]

It's the same characteristics as the 1st clip and using tfm().tdecimate() works fine as suggested in the 1st reply , as it works with all GH1 hard telecined clips (24p in 60i)


Directshowsource("video.mts")
TFM()
TDecimate()

[Guest]

I don't use DirectShow or ffdshow so I can't comment on that.

Both clips are hard 3:2 pulldown. This works fine for me on your second one:

loadplugin("dgdecodenv.dll")
dgsource("60i_gop5_.dgi")
assumetff()
telecide()
decimate()

It's just simple hard 3:2 pulldown with AVC video.

[ExperimentalAnimations]

Thanks for the help but it doesnt do me any good if vdub is automatically IVTC'ing my footage.

[poisondeathray]

Ive also checked my directshow filters (ffdshow) to see if its doing processing and everything is unchecked. Graphedit is showing my decompressor as just "avi decompressor" with little "xforms in/outs" on it, which goes out to VMR video renderer. Unless you see any problems with that i dont know what else to do.


render the .mts in graphedit / graphstudio, not the .avs

do you have a directshow splitter , like haali media splitter installed?

can you play the .mts files in mpc ?

Thanks for the help but it doesnt do me any good if vdub is automatically IVTC'ing my footage.

vdub doesn't do anything automatically unless you tell it to

maybe your directshow system is applying some filters

[ExperimentalAnimations]

Put dgdecode.dll in avisynth 2.5.7 plugin folder and loaded the script in vdub and it gives me "unable to load "dgdecodeenv.dll". Apparently its a paid program. Also tried dgdecode.dll, still doesnt work.

[Guest]

Do what poisondeathray said: Open the MTS in GraphEdit, not the AVS. Report what happens.

I gave my script just to show that it is simple hard 3:2 pulldown that is easily handled.

[ExperimentalAnimations]

Ah yes, turns out to be directshow was set to deinterlace and whatnot by default. I thought graphedit would have picked up on that through the script.

Tried the first script and it works like it did before, but im getting combing artifacts again, and this time their really ugly.

http://h.imagehost.org/t/0410/artifacts.jpg (http://h.imagehost.org/view/0410/artifacts)

[Guest]

I don't see any combing in that image. Where is it?

I do see what looks like artifacts from a postprocessing deinterlacer. MY script does not produce that so it is still something in your DirectShow filter or script.

So please post the current script so I can see what you are doing. And what DirectShow filter is being used?

[ExperimentalAnimations]

Directshowsource("..")
assumetff()
TFM()
Tdecimate()

Its not always there, its just on certain frames. In the image theres combing patterns on the wall. Look at the shadows at the top.

[Guest]

There's no combing there. Do you know what combing is? Put a circle on the image where you see combing.

This is an interesting case because it suggests that Telecide()/Decimate() defaults work better than TFM()/TDecimate() defaults for this clip.

[poisondeathray]

Answer the questions about the decoder and graphedit results

I see the artifacts in your screenshot, but I don't them when I preview the avs script on my system

Are you scrubbing back & forth in vdub ? If you use DirectShowSource() it can give artifacts (frames get misordered), but should be ok if you do a straight linear encode.

Another option is to use DSS2() and Audiodub if you want audio with the clip , or DGNVTools to index the clip

[Guest]

I see the artifacts in your screenshot You see combing in that image? Where? Or do you mean the crappiness at the top edge of the vase for example?

[poisondeathray]

You see combing in that image? Where? Or do you mean the crappiness at the top edge of the vase for example?


I said I saw artifacts, but I'm not certain if you can call it residual combing

They look a bit reminiscent of yadif deinterlacing artifacts (e.g. around the jar) , and there might be some combing around the table edge

Whatever they are, I can't reproduce those artifacts here with that script (his screenshot indicates frame #173 , but I looked all around and can't see them even on other surrounding frames)

For example, here is frame 173 , and it definitely does not have those artifacts in his screenshot, and it looks much cleaner (Same script, haali media splitter, ffdshow decoder)
http://i54.tinypic.com/j9alpl.png

[ExperimentalAnimations]

Ok, i guess its not like gigantor interlace combing, its superfine and splotchy.

http://h.imagehost.org/t/0931/artifacts_LOOK.jpg (http://h.imagehost.org/view/0931/artifacts_LOOK)

As for the jar, i dont know what that is, looks like the image tore or something.

[poisondeathray]

^ Have a look at the same frame I posted above , there are no artifacts like there are in your screenshot

The artifacts I'm referring to are dot like artifacts on the periphery of the jar , and the table edges in your screenshot

[Guest]

I said I saw artifacts, but I'm not certain if you can call it residual combing

They look a bit reminiscent of yadif deinterlacing artifacts (e.g. around the jar) , and there might be some combing around the table edge

Whatever they are, I can't reproduce those artifacts here with that script (his screenshot indicates frame #173 , but I looked all around and can't see them even on other surrounding frames)

For example, here is frame 173 , and it definitely does not have those artifacts in his screenshot, and it looks much cleaner (Same script, haali media splitter, ffdshow decoder)
http://i54.tinypic.com/j9alpl.png OK, sure. My script doesn't make them either so we are in agreement.

Has he told us yet what DS filter he is using?

"Answer the questions about the decoder and graphedit results"

[poisondeathray]

Has he told us yet what DS filter he is using?

not yet... it feels like pulling teeth to get information here...

it could go a lot faster if the questions asked were answered :)

[ExperimentalAnimations]

Divx H.264 decoder is what its currently set to. What else do you want to know? Ive doubled checked it, theres no processessing going on.

[poisondeathray]

you're certain ? when you render the .mts file in graphedit, are there any other filters ?

what about vdub filters?

what happens if you enable ffdshow for h264/avc decoding ? (when you do that the merit should automatically be higher than divx h264, but you can double check in graphedit)

can you answer the previous question about scrubbing (seeking around in vdub) ?

[Guest]

not yet... it feels like pulling teeth to get information here...

it could go a lot faster if the questions asked were answered :) I didn't want to say that to avoid being accused of harrassment of noobs. :)

[ExperimentalAnimations]

Thats weird, its gone, but then it comes back again in different places. It might be a vdub bug.

[ExperimentalAnimations]

Ive tried jumping around, scrubbing backwards, etc... I cant figure what exactly is causing it.

No other filters in graphedit, no filter in vdub, tried different processing setting, using ffdshow didnt have any effect.

[Guest]

I believe poisondeathray attributed it to DirecShowSource() random access issues. That is likely the cause, I agree.

Use a source filter that supports reliable random access. DSS2() may work here, it's worth a try.

Or just encode linearly from the start and see what happens.

[ExperimentalAnimations]

Oh, i tried DSS2 but apparently i dont have it. Link plz?

[Guest]

Install Haali Media splitter. Then in the install directory you should find avss.dll. Load that in your script.

[2Bdecided]

Last time I saw this it was a codec issue - using a codec that was happy with the stream solved it.

You can get all kinds of random junk if you try to seek with directshowsource.

Combine these two problems, and it's to be expected that you always get some kind of problem, but rarely the exact same problem twice in a predictable way.

(I haven't found a universally working free decoder for h.264 unfortunately. I've no reason to throw money at the problem yet, and don't have an NV graphics card.)

Cheers,
David.