Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > MPEG-4 ASP

Reply
 
Thread Tools Search this Thread Display Modes
Old 19th February 2003, 22:48   #1  |  Link
SnotRat
Registered User
 
Join Date: Feb 2003
Posts: 24
Help. What do I do with this.. Force Film, IVTC?

I'm trying to encode Return to Oz NTSC. DVD2AVI shows me film less than 95% and it fluctuates between progressive and interlaced. I tried force film first which resulted in artifacts. Then i used field operation none, and when looking at it frame by frame in Gknot i see two interlaced frames out of five consistently throughout which i read is the criteria for IVTC. So I tried it with IVTC and there are still few small parts with interlace artifacts. The only way i got it void of artifact is by selecting Force Film and then Field Deinterlace. Should I stick with the IVTC and deal with the little bit of artifact or should I Force film and Field Deinterlace? Or is there a better way entirely?

Last edited by SnotRat; 19th February 2003 at 23:44.
SnotRat is offline   Reply With Quote
Old 19th February 2003, 23:46   #2  |  Link
jggimi
Moderator Emeritus
 
jggimi's Avatar
 
Join Date: Dec 2001
Location: The NW corner of Lake Erie
Posts: 5,552
If you're using IVTC via Gknot, you are using DeComb's Telecide and Decimate filters. Telecide includes Field Deinterlacing by default, so your symptoms are a little confusing to me.

I do not recommend using Force FILM in this case, as the MPEG-2 flags are not set up for it.

See this and this, then...see the discussion with neuron2 and manono on pattern matching in this.

Edit: 3rd link corrected

Last edited by jggimi; 20th February 2003 at 01:21.
jggimi is offline   Reply With Quote
Old 20th February 2003, 00:45   #3  |  Link
SnotRat
Registered User
 
Join Date: Feb 2003
Posts: 24
I've read through the IVTC faq and from my understanding this movie is the perfect canidate for IVTC which leaves me wondering why i still ended up with some artifacts.

Whether i use Telecide() or Telecide(post=false) the results are the same. But I've narrowed it down to only one place where the artifacts occur after IVTC. I opened the .d2v in gknot to that exact place thinking that there might be more interlaced frames than 2 out of 5 in that spot, but there weren't. It stayed at 3 frames non-interlaced followed by 2 interlaced consistently.

The sequence where this happens only lasts about three seconds and i notice about five frames with artifacts. Maybe i should just ignore it? I would just like to better understand why this happened.
SnotRat is offline   Reply With Quote
Old 20th February 2003, 01:20   #4  |  Link
jggimi
Moderator Emeritus
 
jggimi's Avatar
 
Join Date: Dec 2001
Location: The NW corner of Lake Erie
Posts: 5,552
That third link was invalid. Try this one for a discussion of pattern matching and/or thresholds...

http://forum.doom9.org/showthread.php?s=&threadid=45266
jggimi is offline   Reply With Quote
Old 20th February 2003, 09:01   #5  |  Link
SnotRat
Registered User
 
Join Date: Feb 2003
Posts: 24
I'm losing my mind. I thought i was making progress but very strange things are happening to me. Like i'll open an avs with IVTC(44,11,95) in Virtual Dub and it'll look fine. But then i'll open the same unchanged avs in VDub five minutes later and there will be interlace artifacts in the exact same frames that appeared fine the previous time i checked it.
SnotRat is offline   Reply With Quote
Old 20th February 2003, 11:54   #6  |  Link
manono
Moderator
 
Join Date: Oct 2001
Location: Hawaii
Posts: 7,406
Hi-

You might try DJ Bobo's newly discovered settings:

IVTC(70,10,25)

He thinks they're better than the previous ones, and I agree. Also, sometimes you have to kind of "sneak up" on the troublesome frames. That is, it has to discover the pattern first, so just stopping right on the frames you want won't always work. You start 10 or so frames before the ones you really want to check, and advance on them frame by frame.

But even IVTC2.2 isn't perfect, and it's often a good idea to load the Decomb Plugin and use FieldDeinterlace() in the script.
manono is offline   Reply With Quote
Old 21st February 2003, 00:27   #7  |  Link
SnotRat
Registered User
 
Join Date: Feb 2003
Posts: 24
This is really confusing and I have many questions. I thought the reason for using ivtc was so you don't have to use FieldDeinterlace() which reduces overall quality. What would be the difference between using Force Film and FieldDeinterlace vs IVTC(70,10,25) and FieldDeinterlace? If I got less interlaced frames when using Force Film than I did IVTC would that indicate that I should use Force Film or assume the IVTC i used didn't work right and try another method or setting?

Now I have a source that when looking at a d2v with field operation set to none shows me 2 out of 5 interlaced frames consistently. I've tried Telecide().Decimate(5) which didn't work in this case. I've read the guides and still don't fully understand decimate(5). I take it to mean it deletes 1 frame out of 5. if i have 2 interlaced frames out of 5, doesn't that mean there will still be 1 interlaced frame remaining per cycle?

Ok. Even with a near complete lack of understanding I managed to get some good results. I was reading through the decomb guide and it made mention of "NTSC with 3:2 pulldown you can enable pattern guidance by setting Telecide(guide=1)" which i reasoned i might have since i noticed 3 progressive frames followed by two interlaced. I don't know if this is correct or not but I used Telecide(Post=false,Guide=1) and Decimate(5) and it appears to have worked.

I also had good luck with both IVTC(70,10,25) and IVTC(44,11,95). I don't know which is better to use if they're both giving me similar results. Or if I should use (Post=false,Guide=1) Decimate(5).

Sorry for all the newb questions but I've read through the guides and a lot of it still looks like hieroglyphics to me.
SnotRat is offline   Reply With Quote
Old 21st February 2003, 01:40   #8  |  Link
N_F
Do you find me... pretty?
 
N_F's Avatar
 
Join Date: Apr 2002
Location: Sweden
Posts: 760
That's a lot of questions... I'll take a few quick stabs (I'm no expert on the area so I'm sure someone will correct me on some points )

I thought the reason for using ivtc was so you don't have to use FieldDeinterlace() which reduces overall quality.

Both yes and no. IVTC (InVerse TeleCine) is a process were you remove some blended fields to get a nice and clean progressive picture. Basically you take a source with a framerate of 29,97 and some ugly blended fields (that would be the 2 interlaced frames you'll see every 5 frames) and perform an IVTC to end up with a framerate of 23,976 where the blended fields have been discarded. This will save a LOT of bits and you'll end up with much better quality.

To summon it up:
A IVTC operation should give you a nice progressive picture at 23,976
Fielddeinterlace() should give you a bit worse looking picture at 29,97


So why did manono recommend IVTC2.2 followed by Fielddeinterlace()? I've never used IVTC2.2, but the general idea should be that IVTC2.2 performs an IVTC, but it may still leave some interlaced artifacts. This is where Fielddeinterlace steps in to (hopefully) remove these artifacts, and Fielddeinterlace is smart so it only deinterlaces frames where it has discovered interlaced artifacts.


What would be the difference between using Force Film and FieldDeinterlace vs IVTC(70,10,25) and FieldDeinterlace? If I got less interlaced frames when using Force Film than I did IVTC would that indicate that I should use Force Film or assume the IVTC i used didn't work right and try another method or setting?

As I said I've never used IVTC2.2, but I would suspect you're right when you say "assume the IVTC i used didn't work right". I have no technical aspects to stand on, but I would think Force Film can be better than a "bad" IVTC (please correct me if I'm wrong on this).


I've read the guides and still don't fully understand decimate(5). I take it to mean it deletes 1 frame out of 5. if i have 2 interlaced frames out of 5, doesn't that mean there will still be 1 interlaced frame remaining per cycle?

You musn't forget the telecide() when you speak of decimate(). These are the basics:

1. Telecide() will check through the frames and rearrange the fields where needed. So in your normal movie where in every 5 frames you have 3 progressive and 2 interlaced (the correct term in this case is really 'combed', not interlaced) telecide() will take care of these 2 interlaced frames. However, these two frames will be identical and this is where Decimate() comes in.

2. If you use decimate(5) it will check through 5 frames and remove 1 of these duplicates that telecide() created. It's in this process the movie is transformed from 29,97 to 23,976 fps (if you'll do the math you'll see it all adds up).


I don't know if this is correct or not but I used Telecide(Post=false,Guide=1) and Decimate(5) and it appears to have worked.

If you're sure it worked than by all means use it. Normally however it's quite strongly recommended to use post=true. This may produce identical results, but it may also take care of some few interlaced artifacts that slip through (which seemed to be your original problem). Normally the only reason to use post=false is that it's a lot faster.



I'm not in the habit of writing long posts like this (I leave that up to monono, hakko504 and jgimmi ), so take it for what it is, no more. I may be plain wrong on some points...

Last edited by N_F; 21st February 2003 at 01:43.
N_F is offline   Reply With Quote
Old 21st February 2003, 03:42   #9  |  Link
manono
Moderator
 
Join Date: Oct 2001
Location: Hawaii
Posts: 7,406
Pretty good explanation, N_F.

The only thing I would quibble with is this:

IVTC (InVerse TeleCine) is a process were you remove some blended fields

But you corrected yourself later when you started talking about how Telecide rebuilt the frames and Decimate pulled out the duplicates. Blended fields are something quite different. Although the IVTC process can have a side benefit of removing blended fields in some cases, SnotRat is talking about getting rid of that 3 progressive, 2 interlaced frame pattern and in the process converting from 29.97fps to 23.976fps.

There are a couple of things I would like to add. One is something you mentioned also, N_F. Telecide has the deinterlacer built in by default (which SnotRat disabled with Post=False). Decomb's deinterlacer doesn't deinterlace all frames (unless you tell it to), but only the ones detected as combed. So, unlike other deinterlacers such as VerticalReduceBy2 or GKnot's SeparateFields().SelectEven() which do deinterlace everything, and can compromise quality, Telecide's deinterlacer shouldn't mess up the rest of the video.

The reason I mentioned adding FieldDeinterlace to IVTC2.2 is because it doesn't have a deinterlacer built into it the way Decomb does. Unless you're real sure that no interlaced frames have slipped by whichever IVTC you use, then you should add it to the script. For me, there's nothing worse than watching the movie later and finding a few interlaced frames. Yes, it slows down the encoding, but it's good insurance.

What would be the difference between using Force Film and FieldDeinterlace vs IVTC(70,10,25) and FieldDeinterlace?

There are 2 reasons why it's said to not Force Film when the percentage is less than 95% in DVD2AVI (and sometimes even if the percentage is greater than 95%). One is that interlaced frames may be slipping through, and the other is that the video may play jerky in places. Sometimes Force Film combined with FieldDeinterlace might work, but I've never used it that way, and don't intend to. I'll always IVTC when in doubt.

And as you discovered, Guide=1 is a good idea when using the Telecide-Decimate combination for NTSC DVD conversions.

Although I prefer Decomb, others use IVTC2.2, particularly for anime. If DJ Bobo says that IVTC(70,10,25) is better than IVTC(44,11,95), then that's good enough for me, as he's the resident expert on it around here. In my testing, I've also found cases where IVTC(70,10,25) did a better job than the other setting.

Good job N_F. You've come a long way, particularly for someone living in a PAL country.
manono is offline   Reply With Quote
Old 21st February 2003, 04:52   #10  |  Link
Guest
Guest
 
Join Date: Jan 2002
Posts: 21,901
My turn to quibble.

When using FieldDeinterlace() as a postprocessor for IVTC22 (and others), you must also set full=false and tweak your threshold if required. Otherwise, it will deinterlace every frame instead of just the combed ones. In Telecide() the postprocessing has full=false by default but in FieldDeinterlace() it is true by default (because standalone FieldDeinterlace() is normally used on interlaced video).
Guest is offline   Reply With Quote
Old 21st February 2003, 05:38   #11  |  Link
Guest
Guest
 
Join Date: Jan 2002
Posts: 21,901
Some babbling about FORCE FILM (assumes DVD2AVI 1.76)...

The only time you can completely safely get away with FORCE FILM is when the numbers in the D2V file consistently cycle like this:

...0 1 2 3 0 1 2 3 0...

That is the 3:2 pulldown signature, shown as "FILM" by DVD2AVI. What can happen legally is that 0's or 2's can repeat an arbitrary number of times. E.g. (now I delete the spaces for brevity):

...01230123000000000000000001230123...
...012301222222222222301230123...

The runs of 0's or 2's are non-3:2 sections, or "NTSC" as DVD2AVI will show it. What can these sections be? Well, just about anything! Here are some of the things I've identified in them:

1. Static sequences where just duplicating 1 frame in 4 is done to get to 30fps, instead of duplicating 2 fields in 10 (it works because the scene is static). This is very common.

2. 30 fps content (interlaced with each field a new picture). Special effects, animations, credits, video, etc.

3. "Hard" 3:2 pulldown. 3:2 pulled down field pattern but encoded as interlaced (pulldown was applied prior to MPEG encoding). Often you have a complete movie that is hard pulldowned; it can be IVTC'ed after doing non-FORCE FILM, but cannot be FORCE FILM'ed.

4. Little violations of the repeat flag setting. The 3:2 field pattern continues but the repeat flags are missing (1 and 3 carry the repeat flag enabled).

5. Probably other cases also, and all combinations of these cases mixed in one overall clip/movie!

If your D2V shows consistent 01230123 cycling with 0 and 2 runs only for static scenes, then you can safely use FORCE FILM. That is the source of the traditional wisdom that if DVD2AVI shows 95% FILM or more you can get away with FORCE FILM.

If you have any 0 and 2 runs that aren't just static scenes, then you will likely end up with combed frames if you use FORCE FILM. Yes, they can be cleaned up by postprocessing with FieldDeinterlace(), but some if not all of them may actually be progressive frames that can be properly restored with Telecide() (if you do not use FORCE FILM).

Finally, using FORCE FILM and postprocessing with FieldDeinterlace() forces you to accept fixed, dumb decimation of the 0 and 2 runs. Decimate() generally does a much better job.

I'm sorry this can't be made easy for the newbies. It is a complex thing. Hopefully, this gives a better understanding of some of the conventional wisdom about using FORCE FILM or not.

@manono

If you think that would be useful in your tutorial, feel free to use it.

Last edited by Guest; 21st February 2003 at 05:53.
Guest is offline   Reply With Quote
Old 21st February 2003, 06:47   #12  |  Link
jggimi
Moderator Emeritus
 
jggimi's Avatar
 
Join Date: Dec 2001
Location: The NW corner of Lake Erie
Posts: 5,552
And that was an absolutely fabulous description, not only of the flags and DVD2AVI's interpretation, but also, how to examine the flags manually in the .d2v project.

Thank you!
jggimi is offline   Reply With Quote
Old 21st February 2003, 12:25   #13  |  Link
manono
Moderator
 
Join Date: Oct 2001
Location: Hawaii
Posts: 7,406
Awesome!
manono is offline   Reply With Quote
Old 22nd February 2003, 01:34   #14  |  Link
N_F
Do you find me... pretty?
 
N_F's Avatar
 
Join Date: Apr 2002
Location: Sweden
Posts: 760
Quote:
Originally posted by manono
Good job N_F. You've come a long way, particularly for someone living in a PAL country.
Thanks. "If I have seen far it's because I've stood on the shoulders of giants", or what was it he said?


@neuron2

At first glance that seems to go a bit over my head, but it does indeed seem interesting. I'll have to read it more carefully as soon as I have the time.
N_F is offline   Reply With Quote
Old 27th February 2003, 09:33   #15  |  Link
kxy
Registered User
 
Join Date: Oct 2001
Posts: 343
This is a NTSC Simpsons
7 0 7B96C 2 2 2 2 2 2 2 2 2
7 0 7BA1F 2 2 2 2 2 2 2 2 2 2 2 2
7 0 7BB09 2 2 2 2 2 2 2 2 2 2 2 2 2 2
7 0 7BBF9 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
7 0 7BCF4 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
7 0 7BDFC 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
7 0 7BEE4 2 2 2 2 2 2 2 2 2
7 0 7BFA9 2 3 0 1 2 3 0 1 2 3 0 1
7 0 7C078 2 3 0 1 2 3 0 1 2 3
7 0 7C142 0 1 2 3 0 1 2 3 0 1 2 3
7 0 7C21D 0 1 2 3 0 1
7 0 7C2A4 2 3 0 1 2 3 0 1
7 0 7C359 2 3 0 1 2 3 0 1 2 3 0 1
7 0 7C43F 2 3 0 1 2 3 0 1 2 3 0 1
7 0 7C528 2 3 0 1 2 3 0 1 2 3 0 1
7 0 7C615 2 3 0 1 2 3 0 1 2 3 0 1
7 0 7C6EB 2 3 0 1 2 3 0 1 2 3 0 1
7 0 7C7DD 2 3 0 1 2 3 0 1 2 3 0 1

I use
Code:
 
Telecide(guide=1,gthresh=50,chroma=true,threshold=30).Decimate(5)
Dup()
It produces the following bad results
image 1 was superimposed on 2
image 3 has jagged edge
The next frame it went back to normal.

image 1.

Last edited by kxy; 27th February 2003 at 09:37.
kxy is offline   Reply With Quote
Old 27th February 2003, 09:35   #16  |  Link
kxy
Registered User
 
Join Date: Oct 2001
Posts: 343
image 2
kxy is offline   Reply With Quote
Old 27th February 2003, 09:36   #17  |  Link
kxy
Registered User
 
Join Date: Oct 2001
Posts: 343
image 3
kxy is offline   Reply With Quote
Old 27th February 2003, 12:19   #18  |  Link
Guest
Guest
 
Join Date: Jan 2002
Posts: 21,901
Your gthresh is rather high.

The image superpositions are caused by blended fields in your source. You can prove it with SeparateFields().

Field matching is not perfect. If you develop a perfect algorithm, please let us all know. It can also be caused by problems in your source as described in the help file. If you are getting only one or two failures, it is nothing to agonise about. You can try manual overrides if you are really bothered about it.
Guest is offline   Reply With Quote
Old 27th February 2003, 22:39   #19  |  Link
kxy
Registered User
 
Join Date: Oct 2001
Posts: 343
Let me appologize. I didn't know the image will turn out to be that big, I did a preview post but couldn't see the image. Mod if you want to edit the picture or make it smaller, feel free.

neuron2,
I am just a newbie, besides turn the gthresh down a bit, what else would you suggest me trying? Can you give me some sample codes to get me started?

Thank you for explaing the superimposed effect. What about the jagged edge on frame 3, the source didn't have it and after I IVTCed, it appears every so often.

Offtopic: Hey you are from Chicago, IL. Small world after all. I was born and raised down in Bridgeport.
kxy is offline   Reply With Quote
Old 28th February 2003, 08:19   #20  |  Link
hakko504
Remember Rule One
 
hakko504's Avatar
 
Join Date: Oct 2001
Location: SWEDEN
Posts: 1,611
@kxy

I did consider cropping the pictures, but I wasn't sure of what parts you would liked to keep.
And the reson you didn't see the pictures at first is that they have to be validated by a moderator before they show up in the forum.
__________________
/hakko

http://www.boardgamegeek.com
hakko504 is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 17:15.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.