How can i convert an interlaced vob in a progressive vob without compressing it? [Archive]

SeeMoreDigital

23rd January 2007, 21:18

Yes i know, it's a kind of compression, but i would like just a conversion from interlace to progressive without compressing it more than it actually is. I believe you caught my point.You can't convert interlaced MPEG-2 to progressive MPEG-2 without re-encoding....

That said, if your discover that your MPEG-2 sources are incorrectly flagged ie: progressive MPEG-2 is flagged as interlaced MPEG-2 (or vice-vera), it is possible to correct this!

epaminondas20

24th January 2007, 16:17

encoding is always a lossy operation , to know is how much you want lose...

i can suggest a greedy searching about deinterlacers in these forums (and take a look to these name scharfis_brain and Didee, have done wonderfull deinterlacers/bobbers)

BHH

Thanks.

Well another question, Is there possible ways to manually deinterlace using MeGui or AVS scripts??. Is there a way to manually select the good fields from the bad fields using MeGui (but in a manual method, like the manual ivtc in TMPGEnc)???

buzzqw

24th January 2007, 16:41

Is there possible ways to manually deinterlace using MeGui or AVS scripts??.

Within MeGui there is a very good interlacer detection program and megui can optimally deinterlace most source with very good results

but selecting good/not good fields... i don't know exaclty what you mean... (it's a very long time since i don't use tmpeg...) , i can only say that tivtc is a very good plugin

BHH

P.S. even my automkv can lead to good results :D

epaminondas20

26th January 2007, 23:05

Well thanks for the tip about AutoMKV. My primary question is if you know any plugin in avisynth that will allow me to manually select good from bad fields just as the TMPGEnc. OR, to be more general, i really want to have perfect deinterlacing of my videos. I am not sure if MeGui has a very good interlacer detection program, since 1 or 2 vobs i had the chance to lay eyes on found them progressive when there were some frames that were interlaced. I also have a vob, that i want you to see (will have to upload) that all my players (Media Player Classic, VLC show it as progressive, if for MPC the MPEG-2 filter is set to Weave and for VLC no deinterlacing, and when i analyse it with Megui it says that it is deinterlaced.

Forgive me for my bad english

Please check in this to see for the vob

setarip_old

27th January 2007, 00:03

i really want to have perfect deinterlacing of my videos.Hi!

Try using "MPEGMediator"...

epaminondas20

27th January 2007, 12:19

well thanks, but MPEGMediator is doesn't convert to matroska video as i do. The question still remains if i can manually do that by my self.

buzzqw

27th January 2007, 13:39

in automkv you can manually choose the deinterlacer or write yourself it

look at this post http://forum.doom9.org/showthread.php?p=937978#post937978

i have explained how deinterlacer works within automkv

best regards :)

BHH

epaminondas20

27th January 2007, 14:10

that's a very cool program you have, but isn't there a way it can detect if source is progressive or interlaced or film or whatever?? (like MeGui)

Well i found this site about custom mpeg-pull down. This is what i mean i would like to do with megui. is there a similar way? or of there is a similar way with automkv?? :D

BUZZQW, i am trying AutoMKV right now.

SeeMoreDigital

27th January 2007, 14:47

that's a very cool program you have, but isn't there a way it can detect if source is progressive or interlaced or film or whatever?? (like MeGui)I suggest you run your MPEG-2 sources thru' DGIndex before hand...

epaminondas20

27th January 2007, 16:22

well that's what i do first

buzzqw

27th January 2007, 16:50

that's a very cool program you have, but isn't there a way it can detect if source is progressive or interlaced or film or whatever??

as i wrote BAutoDeint after analysis could decide that source file is: automkv will always analyze the source (if you leave auto on deinterlacer).

if you get some problem , please post :)

BHH

manono

27th January 2007, 17:33

Hi-

Forgive me, but I can't tell from your posts if you actually know anything, or are blissfully ignorant of the subject. Are you PAL or NTSC? If NTSC, are we talking about an IVTC (originally shot with 24fps film cameras) or a true deinterlace (shot with 30fps video cameras)? If PAL, is it really interlaced, or at its heart really progressive, perhaps with shifted fields? How do you know your source is interlaced? Are you getting the information from something like DGIndex (doesn't mean much in this context) or from examining the frames yourself? Is it possible that your source is the victim of a bad standards conversion and has blended fields?

Would you care to post a short 10-15 second sample for us to examine, so that we can give some more educated advice? If you need instructions as to how or where, just ask.

It's been so long since I've used TMPGEnc for deinterlacing (at least 5-6 years), that I forget what the choices are, but TDeint allows you to choose the field from which to interpolate. In addition, by using an override file, you can control the values for as much or as little of the video as you like. And I believe that there are several AviSynth deinterlacers better than anything in TMPGEnc.

reepa

28th January 2007, 06:26

You can't convert interlaced MPEG-2 to progressive MPEG-2 without re-encoding....

That said, if your discover that your MPEG-2 sources are incorrectly flagged ie: progressive MPEG-2 is flagged as interlaced MPEG-2 (or vice-vera), it is possible to correct this!

How do you find out whether progressive or interlaced encoding has been used since you can't rely on the flags? I've tried DGIndex, GSpot, MPEG Parser and Restream, but afaik they all rely on the flags for information. The file I'm examining is an m2v I demuxed from a commercial PAL DVD. The flags claim the video is interlaced and top field first.

manono

28th January 2007, 08:44

DGIndex tells you how it was encoded, and not whether the source is interlaced or not. There's a distinction there I was trying to bring out in my previous post. The vast majority of PAL movies on DVD are encoded as interlaced, but the vast majority are from progressive sources. You have to examine the frames to be sure of what you have, and not let DGIndex, ReStream, Bitrate Viewer, or anything else, tell you.

I hate to disagree with my esteemed colleague SMD, but if it's encoded as interlaced, and flagged as interlaced, then it's not incorrectly flagged. It's perfectly OK to take a progressive source and encode it as interlaced. So, to answer your question, reepa, if DGIndex says it's encoded as interlaced and is TFF, the chances are very good that it's true. There aren't all that many movies on DVD that are incorrectly flagged. A few, to be sure, but not many.

SeeMoreDigital

28th January 2007, 15:50

I hate to disagree with my esteemed colleague SMD, but if it's encoded as interlaced, and flagged as interlaced, then it's not incorrectly flagged. Yes indeed...

I too was referring to correcting the flags of incorrectly flagged sources...

Sufficed to say, DGIndex is "the" perfect tool for analysing the frame type of MPEG-2 video streams. Because unlike media file readers (such as MediaInfo, GSpot, etc) it looks at the actual MPEG-2 video stream and just at the streams header information.... Or use your eyes!

epaminondas20

28th January 2007, 17:02

Well, isn't that obvious that i have posted this topic in the newbie section??

Ok, to get it over with. I would like to make a 15 to 20 second sample of the vob i want to deal with, but what program can do that??

thanks

Guest

28th January 2007, 17:22

Load the VOB in DGIndex. Set a range with the [ and ] buttons. Then do File/Save Project and Demux Video. You will get an M2V file. Upload that.

Or if you want us to see the audio too, then use ChopperXP to make a fragment of the VOB.

reepa

28th January 2007, 18:11

Ah, I was looking at the wrong field on the DGIndex information panel. "Frame Struct" tells whether the video is encoded progressive or interlaced. Frame Struct: Frame means progressive and Frame Struct: Field means interlaced. ReStream and MPEG Parser call it "Picture structure". This cannot be changed without re-encoding. This is the information I was interested in!

Additionally, there are flags that instruct how to output the decoded video, which can be changed without re-encoding:

For every MPEG picture(I, P, B):
* Picture type (progressive or interlaced): DGIndex calls it "Frame Type" (progressive or interlaced); ReStream calls it "Frametype Progressive" (true or false); MPEG Parser calls it "Progressive Frame" (0 or 1)
* Field order (top field first or bottom field first): DGIndex calls it "Field Order" (top or bottom); ReStream calls it "Top field first" (true or false); MPEG Parser calls it "Top field first" (0 or 1)
* Repeat first field (yes or no): only MPEG Parser shows this

For every MPEG sequence (GOP?):
* Sequence type (progressive or interlaced): ReStream and MPEG Parser call it "Progressive Sequence" (true or false)

No wonder people get confused :( For reference, the video I'm examining (Deer Hunter PAL R2 Nordic release) is encoded progressive (Frame Struct: Frame), but flagged interlaced (Frame Type: Interlaced, Field Order: TFF, Repeat First Field: No, Progressive Sequence: No). Ridiculous, IMO. So in conclusion: epaminondas20 can look at the video information with DGIndex. If Frame Struct is Frame, you can easily "convert" the video from interlaced to progressive by changing the flags (with ReStream for example).

Guest

28th January 2007, 18:48

* Repeat first field (yes or no): only MPEG Parser shows this. DGIndex shows the running count of field repeats, so you can single step and see if the count increments.

epaminondas20

28th January 2007, 18:56

Where do you want me to upload that m2v video??? What suits you best?

Guest

28th January 2007, 19:25

I'm not sure why we're doing this, but if you want to upload it to Megaupload.com, we can have a look at it.

epaminondas20

28th January 2007, 20:03

I already uploaded a segment of the video in sendspace, but i will upload a second segment for some extra reference i need to ask for

This is the first video sample that i did, hxxp://www.sendspace.com/file/vzpmp0 (replace xx with tt, while i know probably i can't hotlink in this site). This one after analysis in Megui, tells me that it is progressive.

While i have done another sample that has some of the frames the previous one has but it's from the beginning and the Megui says it's interlaced. Check this one too, hxxp://www.megaupload.com/?d=OX0RPT2F

Anyway, seems like the d2v i get with DGIndex makes a good job

Guest

28th January 2007, 21:33

They're both progressive PAL content but encoded as interlaced.

There should be no problem at all processing this material. What problem are you having or are you inventing ghosts?

epaminondas20

28th January 2007, 21:46

i am not inventing ghosts. Rather i would say i should have given you the whole video. and if you see the whole video when you analyse it with MeGui it will tell you it's interlaced. I had to Ignore Pulldown flags in order to make megui understand it as progressive.

Even though i know a not so good way to just see if what i have is interlaced or not, is to select a Weave option in MPClassic MPEG-2 Decoder filter.

reepa

28th January 2007, 22:08

edited: Never mind!

manono

28th January 2007, 23:13

Open a vob in DGIndex. Use the [ and ] buttons to isolate a small section that has movement/motion/action. 10-12 seconds will be plenty. Then File->Save Project and Demux Video. Upload the resulting M2V (ignore the audio and the D2V) to:

http://rapidshare.de/

and post the link they give when done. They'll take up to 50 MB, but you won't need anywhere near that much, probably just 10-15 MB.

Edit: Better never than late. I didn't realize there was a second page to this thread until after I had posted.

SeeMoreDigital

28th January 2007, 23:37

epaminondas20,

At the end of the day, it ain't the hardest thing in the world to "visually" detect whether a standard-def MPEG-2 source is pure interlaced and/or contains some interlaced fielded frames. You need to look out for horizontal lines... Like this: -

http://img249.imageshack.us/img249/6029/interlacedhc4.jpg

Cheers

manono

28th January 2007, 23:46

"Frame Struct" tells whether the video is encoded progressive or interlaced. Frame Struct: Frame means progressive and Frame Struct: Field means interlaced.
Not true. Interlaced encoding can have a pic_structure of either frame or field. For example, this one from DGIndex has a pic_structure of frame, yet uses interlaced encoding:
http://img413.imageshack.us/img413/8484/dgindexus5.th.jpg (http://img413.imageshack.us/my.php?image=dgindexus5.jpg)

The same video, this time in Bitrate Viewer:
http://img411.imageshack.us/img411/189/bitrateviewertd9.th.jpg (http://img411.imageshack.us/my.php?image=bitrateviewertd9.jpg)

And in ReStream:
http://img253.imageshack.us/img253/1618/restreamez6.th.jpg (http://img253.imageshack.us/my.php?image=restreamez6.jpg)

Therefore:
No wonder people get confused :( For reference, the video I'm examining (Deer Hunter PAL R2 Nordic release) is encoded progressive (Frame Struct: Frame), but flagged interlaced (Frame Type: Interlaced, Field Order: TFF, Repeat First Field: No, Progressive Sequence: No). Ridiculous, IMO.
Not ridiculous at all. Your Deer Hunter is encoded as Interlaced.

reepa

29th January 2007, 00:25

Then how come I can just change the encoding from interlaced to progressive by checking the flag and rewriting the stream? According to CWR03 that would require re-encoding, but that's hardly what ReStream does. I wish links to the standards on the official MPEG website worked...

By interlaced encoding I mean the picture is split into two separate fields. DCT, motion prediction, and whatnot is then performed on the two separate fields instead of a complete frame. Isn't that what Frame Struct means?

PS. I still think it's ridiculous to flag / encode progressive video as interlaced.

manono

29th January 2007, 01:06

Then how come I can just change the encoding from interlaced to progressive by checking the flag and rewriting the stream?

Because all you're doing is changing the flagging, and making it incorrect. If the video really is interlaced (as in SMD's pic above), doing that can wreak havoc with playback by flag reading DVD players outputting to progressive displays, such as computer monitors and HDTVs.

According to CWR03 that would require re-encoding, ...

His was the first response, and at the time it looked like the question was about an interlaced source, and not just interlaced encoding. So, what CWR03 (and later SMD) said was true. If it really is interlaced, then to make it progressive, it'll require an IVTC, a deinterlace, or perhaps just a Telecide or TFM by itself, if it's a PAL DVD with shifted fields. Doing those things requires a reencode.

By interlaced encoding I mean the picture is split into two separate fields.

If you're talking about each field being encoded separately, and the pic_structure being field, then while possible and legal, it's fairly rare these days. I have only a very few DVDs where that happens. Of the readily available software MPEG-2 encoders with which I'm familiar, I believe that Procoder is the only one that can do that. Here, read this:

http://www.hometheaterhifi.com/volume_7_4/dvd-benchmark-part-5-progressive-10-2000.html

About a third of the way down is a section called 3-2 Pulldown. Although about NTSC, it'll give you an idea of the specifics. Only one of the 3 encoding examples shown in there has a pic_structure of field, and as I said, that's pretty rare to see these days. The 3rd example shows a pic_structure of frame with interlaced encoding, fairly common even on NTSC DVDs, especially in the extras, and it predominates in PAL DVDs.

In the last paragraph of that section it says:
We should also mention that there is a decent amount of material encoded with 2-2 pulldown, which has a different flag pattern, and is even more likely to not be marked with the "progressive-frame" flag. In Europe almost every film is encoded with 2-2 pulldown. Their TVs run at 50 fields per second, so the easiest way to encode a 24 fps film is to speed it up slightly to 25 fps and show each frame for 2 fields.

reepa

29th January 2007, 02:18

Obviously flagging truly interlaced video as progressive is wrong, because then you'll have combing artifacts because no proper deinterlacing is done. But if the underlying video is really progressive, changing the flag from interlaced to progressive is perfectly okay (as long as every picture has picture_sturct = frame), right?

I understood interlaced encoding as encoding both fields separately, but looks like interlaced encoding can also use progressive frames. Unfortunately, it looks like HD-DVD inherited this confusion.

BTW, I found a properly (by which I mean everything is correctly flagged progressive) authored commercial DVD in my collection: old Disney cartoons.

edit: I would still call all video "progressively encoded" if it has "picture_struct = frame" even if it's flagged interlaced. Similarly, interlaced encoding means "picture_struct = field". Flags can be changed without re-encoding - the underlying encoding scheme can't.

Guest

29th January 2007, 03:01

I'll try to clear up the confusion here.

If you do DCTs, etc., on fields as reepa described then you have field pictures. If you do it on frames then you have frame pictures. As manono points out, usually frame structure (frame pictures) is done, but rarely field structure is used. The idea is to save bits but often the overhead of more pictures counters the savings by doing field pictures. So it's just not used that often because it is not as good in practice as one might think.

So, if you have frame pictures, the content can still be progressive or interlaced. Now, when you downsample the chroma for YV12, you need to do it one way for interlaced content and one way for progressive. That is what Frame Type in DGIndex is telling you: whether the chroma was downsampled using the progressive or the interlaced algorithm.

Theoretically, interlaced content should use the interlaced algorithm and progressive content should use the progressive algorithm. But due to a historical legacy players didn't always properly follow the flagged Frame Type when upsampling for display, leading to a nasty problem called CUE (Chroma Upsampling Error). Content providers found that it was safer to encode everything as interlaced if the players could not be relied upon to follow the Frame Type flag. The practice has persisted and that is why you will see instances of progressive content downsampled as interlaced.

Finally, changing the frame type flags with Restream makes the stream WRONG, because the flags no longer reflect the way the frames were actually downsampled. I assume it would be highly rare to have a stream where the encoder outright lies, i.e., downsamples one way and then sets the flags the other way.

Guest

29th January 2007, 03:08

if you see the whole video when you analyse it with MeGui it will tell you it's interlaced. I had to Ignore Pulldown flags in order to make megui understand it as progressive. You need to pose this in the MeGui forum. They are always interested in tweaking their interlace detection method. It sounds weird that ignoring pulldown for a PAL stream could have an effect, since PAL doesn't use 3:2 pulldown. Can you preview the video in the latest DGIndex beta (honoring pulldown flags) and see if the field repeats or frame repeats boxes in the Info display get incremented? If so, then there is pulldown present.

Even though i know a not so good way to just see if what i have is interlaced or not, is to select a Weave option in MPClassic MPEG-2 Decoder filter. The most reliable way is to separate the fields using Avisynth and then step through them and see if you get field pictures separated in time. I have described the process many times here.

Guest

29th January 2007, 03:14

Then how come I can just change the encoding from interlaced to progressive by checking the flag and rewriting the stream? According to CWR03 that would require re-encoding, but that's hardly what ReStream does. You're right, just changing the flags does not require reencoding, but it does make the stream incorrect. To change the flags and keep the stream correct, you'd have to reencode to do the chroma downsampling to match the revised flags. (I assume that the stream isn't outright lying in the first place but I suppose it could happen.)

I wish links to the standards on the official MPEG website worked... The links in the Library at my site work fine:

http://neuron2.net

reepa

29th January 2007, 04:19

Ah, I thought there was no downside to "reflaggin", but I was wrong. Thanks for the info :^)

Spc01

29th January 2007, 08:45

That said, if your discover that your MPEG-2 sources are incorrectly flagged ie: progressive MPEG-2 is flagged as interlaced MPEG-2 (or vice-vera), it is possible to correct this!
Yes that's true.
But what if you have MPEG-2 interlaced source?

Anyway interlaced source can not be converted to pure progressive scan.
By pure progressive scan i mean that no pixels are lost ... (full resolution).
If you have interlaced source you allready have half of the full resolution so interlaced video is very loosy by means of resolution.
The only thing that's good about interlaced video is that you have twice as much FPS than you have with progressive source but resolution suffers...

If you ask me i would better like to see full progressive scan (full resolution) with less fps than interlaced video with lower resolution and higher frame rate..
:)

*Another directShow filter that doesn't use flags when decoding video is: Ligos MPEG Video Decoder
If you can get ligos MPEG Video decoder you can use it and visually determine if source is interlaced or progressive (this is how i do it)

SeeMoreDigital

29th January 2007, 10:52

Yes that's true.
But what if you have MPEG-2 interlaced source?If your source is truly interlaced and flagged as interlaced, then everything is correct and you should not change the flag info.

If your source is truly interlaced but flagged as progressive (which I've only seen once), then yes, change the flag info.

Have we about done now?

epaminondas20

29th January 2007, 13:07

Well better you get the whole video

http://www.megaupload.com/?d=6IRAER8V

Link Password : www.mvtorrents.com
Rar Password : barth

Sorry it wasn't uploaded by me, i wouldn't make so many passwords if it would be for you.

My problem is that with this video when i save the project through DGIndex when i ignore pull down flags (right part) and when i honor pull down flags (left part) is showing that is progressive on both ways.

http://serv2.imagehigh.com/imgss/4922187_vid1.JPG

This is how DGIndex shows the video, as progressive, no interlaced lines.

http://serv2.imagehigh.com/imgss/4922189_vid2.JPG

Ok and here is the problem.

When i load the d2v file to Megui, that has been created after i chose to ignore pulldown flags the analysis says it's progressive so no deinterlace is suggested and here's the pic

http://serv2.imagehigh.com/imgss/4922192_vid4.JPG

But if i honor the pulldown flags and load that d2v to megui avisynth creator, the analysis say's it's interlaced and the photo shows:

http://serv2.imagehigh.com/imgss/4922308_vid3.JPG

Forgive me for my stupidity, that's why i posted in the Newbies section. Thanks a lot for the info though.

BTW neuron2, i have search the doom9 forums about how i can manually choose the right fields in an interlaced vob, through avisynth, in order to rip that video in x264 matroska. you said you have posted stuff about that, but my search criteria is only limited to my knowledge. I don't mean i need spoonfeeding, i am jjust asking you (with a pleading manner :p) to just tell me which topics to look for.

Thanks a lot for your help.

SeeMoreDigital

29th January 2007, 13:21

A link to you entire video is not neccessary.... We've seen the frames and the relevent comments have been forwarded.

It sounds as though you would benifit by leaning a bit more about, all the different frame types including how they look and work before using functions within applications you are not sure about what they do.

Cheers

manono

29th January 2007, 14:07

epaminondas20

29th January 2007, 14:40

Hi-

BTW neuron2, i have search the doom9 forums about how i can manually choose the right fields in an interlaced vob, through avisynth...

I'm not neuron2, but it doesn't look to me as if it's really interlaced, but perhaps just has shifted fields with Honor Pulldown Flags (and I have no idea why Honor Pulldown Flags is any different from Ignore Pulldown Flags with a PAL video). You can test by creating an AviSynth script with Telecide(Guide=2,Post=0) alone (from neuron2's Decomb), or TFM(PP=0) alone (from tritical's TIVTC). That should serve to realign the fields and make it progressive again. Much better than deinterlacing.

Yes, telecide(Guide=2, Post=0) works.

A link to you entire video is not neccessary.... We've seen the frames and the relevent comments have been forwarded.

It sounds as though you would benifit by leaning a bit more about, all the different frame types including how they look and work before using functions within applications you are not sure about what they do.

Cheers

I would love to know about the different frame types including how they look. Cause i not only want to work videos like these, but also with videos with pure interlacing.

Thanks

squid_80

29th January 2007, 19:13

(and I have no idea why Honor Pulldown Flags is any different from Ignore Pulldown Flags with a PAL video).
The stream is a bit odd - it starts off with frame type = progressive, bff and colorimetry = smpte 170m, then switches to frame type = interlaced, tff and colorimetry = bt. 709. Then at the end of the clip it switches back again. Attempting to save a .d2v with dgindex 1.4.8 results in a "Field order corrected" message and loading the .d2v into avisynth results in shifted fields (unless ignore pulldown is used). Processing the clip with dgindex 1.4.9 beta 13 works fine and gives properly matched fields.
So there's no repeated frames/fields, it's just an odd stream that neuron2 seems to have fixed handling of already :)