(How to) Inverse Telecine

[Cherbette]

I have a video which "is actually progressive content (23.976 fps) and hard telecined. It should be inverse telecined (IVTC) to recover the original progressive frames." However, I'm not entirely sure how to do this in avisynth. I assume it would be one of these two:

MPEG2Source("D:\myproject.d2v")
Telecide()
Decimate(5)

or

MPEG2Source("D:\myproject.d2v")
Telecide()
Decimate(mode=1)


and then I have to decomb or....??

Video: http://www.mediafire.com/?lq06kj4sk6zvich

[jagabo]

For normal 3:2 pulldown material:

MPEG2Source("D:\myproject.d2v")
TFM(d2v="D:\myproject.d2v")
TDecimate()

If it still has comb artifacts (I haven't seen your video sample yet) you need to look closer at what you really have. Make sure Honor Pulldown Flags was used in DgIndex. If DgIndex shows the video is almost all film (99+%) you can just use the Force Film settings and skip the IVTC.

[Cherbette]

It will actually be an AVISource...I forgot to adjust that in the two sample scripts I posted. So DgIndex won't come into play I don't think. But I'm going to try that out on my video. Thank you I'll post some results for anyone who is following

[Cherbette]

Script:
LoadVirtualDubPlugin ("C:\Program Files\VirtualDub\plugins\CCD.VDF", "CCD", 0)
AviSource("Commercial.avi").AssumeBFF().Trim(308,1 179)
TFM()
TDecimate()
ConvertToRGB32(matrix="Rec601")
CCD()
#Crop(5,4,-14,-5)

Progress: http://www.mediafire.com/?gdufdaiaxqus250

[jagabo]

Adding AssumeTFF() before TFM() might help a bit. As usual you have terrible horizontal jitter problems.

[Cherbette]

Oh I forgot...I think this video may have been recorded before I started using the DVD-R as a passthrough for TBC. So I shall recapture

[jmac698]

I've seen this commerical before, didn't you mention it in another thread? Someone made it look better already.

[Cherbette]

Yes, it was mentioned briefly in another thread however I've been working on other videos and never addressed the IVTC issue present in the commercial.

[sanlyn]

I've seen this commerical before, didn't you mention it in another thread? Someone made it look better already.

Well, sort of. But not very much.
https://forum.videohelp.com/threads/339411-Tweaking-VHS-captures?p=2110091&viewfull=1#post2110091

This one's a project in itself, scene by scene. Telecine would be only a small part of the problem.

[sanlyn]

[QUOTE=Cherbette;2121078]Script:
LoadVirtualDubPlugin ("C:\Program Files\VirtualDub\plugins\CCD.VDF", "CCD", 0)
AviSource("Commercial.avi").AssumeBFF().Trim(308,1179)
TFM()
TDecimate()
ConvertToRGB32(matrix="Rec601")
CCD()
#Crop(5,4,-14,-5)
/QUOTE]

This video isn't BFF.

[Cherbette]

Oh...Avisource()
Info()

gave a result of "bottom frame first" so I went with that.

[jmac698]

Hint, mpg2 is usually TFF and captured avi is usually BFF. So a hardware compression capture card would be TFF. But that's just a rule of thumb, the way to tell is separatefields and step through to see if motion keeps going forward or if it jumps back and forth. If it's smooth, your setting is right. Avisynth defaults to BFF I belive.

[Gavino]

Avisource()
Info()

gave a result of "bottom frame first" so I went with that.

That tells you nothing - it's just the Avisynth default.
AviSource can't tell if the video is TFF or BFF (or even if it's interlaced).
You need to examine it visually as jmac698 describes.

[dphirschler]

This is my AviSynth function for IVTC:

# Slick little function which manually reverses 3:2 pulldown.
#
# if p=0, then 4rd and 5th frames are interlaced (pppii)
# if p=1, then 3rd and 4th frames are interlaced (ppiip)
# if p=2, then 2nd and 3rd frames are interlaced (piipp)
# if p=3, then 1st and 2nd frames are interlaced (iippp)
# if p=4, then 1st and 5th frames are interlaced (ipppi)
#
# Parameters:
# clip v  NTSC telecined (3:2 pulldown) video.
# int p   Should be a number 0-4.  If that doesn't work, then try
#         complementParity(v).
function manualIVTC(clip v, int p)
{
    v=loop(v, p, 0,0)# adds extra frames to start pattern
    v=v.SeparateFields().SelectEvery(10, 0,1, 2,3, 5,6, 7,8).Weave()
    v=trim(v,int(p*3/5),0)# deletes the extra frames added
    
    return v.assumeFPS(23.976)
}

To use it, just do "v=v.manualIVTC(x)" where x=0-4. If the pattern changes, then cut it into sections and do the manualIVTC function on each section. Sounds harder than it is. The results are just perfect.


Darryl

[jmac698]

There's an easier way I think,

function simpleivtc(clip in, int phase) {
#IVTC for a fixed 3:2 pulldown pattern, non-adaptive, returns 4/5 framerate clip (i.e. 24fps from 30fps)
in.swapfields.doubleweave
pulldown(0+phase,3+phase)
}

[Gavino]

# Slick little function which manually reverses 3:2 pulldown.
#
# if p=0, then 4rd and 5th frames are interlaced (pppii)
# if p=1, then 3rd and 4th frames are interlaced (ppiip)
# if p=2, then 2nd and 3rd frames are interlaced (piipp)
# if p=3, then 1st and 2nd frames are interlaced (iippp)
# if p=4, then 1st and 5th frames are interlaced (ipppi)
#
# Parameters:
# clip v  NTSC telecined (3:2 pulldown) video.
# int p   Should be a number 0-4.  If that doesn't work, then try
#         complementParity(v).
function manualIVTC(clip v, int p)
{
    v=loop(v, p, 0,0)# adds extra frames to start pattern
    v=v.SeparateFields().SelectEvery(10, 0,1, 2,3, 5,6, 7,8).Weave()
    v=trim(v,int(p*3/5),0)# deletes the extra frames added
    
    return v.assumeFPS(23.976)
}

To use it, just do "v=v.manualIVTC(x)" where x=0-4.

Your code does not work for p=0 - it deletes the first frame (because of the call to loop()).
You need to call it with p=5 for the case pppii.

If the pattern changes, then cut it into sections and do the manualIVTC function on each section.

That's the Achilles heel of any 'manual' method - some sources have many pattern changes due to editing and it is very tedious to find them and cut into sections manually. That's why more 'intelligent' methods such as TFM().TDecimate() are needed.

[dphirschler]

That's the Achilles heel of any 'manual' method - some sources have many pattern changes due to editing and it is very tedious to find them and cut into sections manually. That's why more 'intelligent' methods such as TFM().TDecimate() are needed.

Depends on the show. A lot of times there is no pattern change, or if there is one it's only at the commercial breaks. But I'll admit, sometimes I encounter a show which has pattern breaks every few seconds. Those are not worth it. But every time I use an "automatic" method, combed frames appear. It's usually the scenes where little is moving except for the mouths.


Darryl