DVD-RB and Interlaced Episodic DVDs

[Pridebowl]

Hi,

I'm currently trying DVD-RB 0.93 on "That 70's show - 1st Season" boxset.
After a first try with default settings and CCE 2.67.0027 SP, I found a lot of "interlaced" artifacts on the tv result, which is a little deceiving for the 30 hours of encoding I went trough for each DVDs ... I used default CCE settings, 8 passes and Undot().Deen().

After I saw the result, I figured that maybe Undot().Deen() was not to be used on interlaced source since it is most probably using spacial algorigthms, but after a few tests (using UnfoldFieldsVertical and FoldFieldsVertical and not using them), I do believe that those filters works fine on interlaced sources.

Now, I did some encoding tests with CCE VBR One Pass on DeInterlaced souce, comparing it to Interlaced source and it seems that I can get better compression results on the DeInterlaced source...

Finally, I'm going to try to use Telecide() to deinterlace my source and try to encode everything as progressive frames and figure out if the result has a better look on the tv.

Has anyone tried this yet ?

Thank you.

[TheSeeker]

I would not use deen() on interlaced sources as its a spatial cleaner/smoother, and I believe that is a no-no. Undot() is temporal however so that should be fine. Personally. If were you, I would use Canopus Procoder 2.0. It works so much better on nasty interlaced sources than CCE does. For example, the output from Procoder blew away the output from CCE when I tried encoding Friends seasons 1-2. You shouldn't really even need undot() if you used procoder. But it wouldn't really hurt anything too much.

Otherwise if you dont have Procoder 2.0 and cant get it, then I would try finding some other tweakable temporal cleaners/smoothers. Oh, and 8 passes is severe overkill. I would go with 3, maximum.

AVISynth Plugins:

http://www.avisynth.org/warpenterprises/

[Pridebowl]

Thank you for you awnser, I will give a try to Canopus Procoder 2.0 for episodic discs. However, the preliminary tests I've made with Telecide() deinterlace and Undot().Convolution3d("movieHQ") were good on the TV-Out.

Thanks for the overkill tip, I found that 8 passes was a little too much (!) after reading a little bit more in the forums. I did a lot of research about the episodic encodes but I couldn't find anything that was recent and that used the latest versions of DVD-RB.

Now, if I use Canopus, should the D2V file be in 23.976 fps or 29.970 fps ? I believe that DVD-RB automatically chooses the "Forced Film" switch in DGIndex so I'm not so sure. What is the process used by DVD-RB to convert 23.976 interlaced .m2v encodes to 29.970 VOBs on the third phase ?

Regards.

[writersblock29]

@Pridebowl

A qoute by wmansir from this thread:
http://forum.doom9.org/showthread.php?t=79492&highlight=forced+film

"You don't have to worry about ForceFilm or %NTSC or any of that stuff. DVD-RB recreates the structure exactly as the original. To do this it uses a special .d2v mode that isn't like having FF on or off like a normal .d2v file. This is a custom mode that jdobbs wrote, although I recall Donald Graft saying he might make it available in future versions of DGDecode."

Hope that helps shed a little light on the subject for ya. ;)

[TheSeeker]

Oh and Pridebowl,

The general rule of thumb is to almost never deinterlace. The logic being that the hardware deinterlacer in either your dvd player or your tv (whichever one ends up performing the operation) will do a better job of it then a fielddeinterlace() or a telecide() or even a smart bobber. You obviously seem to know what your doing in regards to filters, but Im curious what made you decide to try telecining these episodes? There are very few instances of dvd's that actually benefit from telecining (at least in Region 1, ntsc dvd's). And incorrect telecining can have serious detrimental effects on the video stream.

[Pridebowl]

The reason why I was looking for an alternative way of encoding the episodic discs is because I had so-so result on the final display and I was able to see a lot of "interlaced artifacts" with a standalone DVD Player on a 40 inches widescreen tv... Probably because CCE is bad for encoding Interlaced (which I didn't know) and because I used Deen which is using spacial algorithms to soften the image (which I was unsure about...).

Also, when you have 3 hours and more of video data on a DVD-9, it is preferable to apply some filters to enhance the compression ratio or you will end up with bad re-encodes... Interlaced sources are much more difficult to handle since you have to make sure your filters are temporal and not spacial, or you must fold the fields and unfold them to apply spacial filters (which I found out how just recently...).

I always thought like you : that sources that were interlaced should stay interlaced but I always had problems with that kind of sources when it went to DivX encoding or DVD Re-Authoring... so it is also a personal choice.

Another thing, I often use my laptop S-Video out on the television to watch DivX or DVD movies and it doesn't properly sync with the TV for Interlaced material, I can often see interlaced lines... but probably there is a way to fix that which I haven't check yet !

I tried a couple of deinterlace filters to find out that Telecide(1) does a pretty good job by zooming 400% with virtualdub so I don't believe that will affect the output in that particular case.

Here is the Bitrate viewer info from the DVD source:

Stream type: MPEG-2 MP@ML VBR
Resolution: 720*480
Aspect ratio: 4:3 Generic
Framerate: 29.97
Nom. bitrate: 9800000 Bit/Sec
VBV buffer size: 112
Constrained param. flag: No
Chroma format: 4:2:0
DCT precision: 9
Pic. structure: Frame
Field topfirst: Yes
DCT type: Field
Quantscale: Linear
Scan type: ZigZag
Frame type: Interlaced
Notes:


I am currently running the Canopus ProCoder 2.0 encode on one machine and the Deinterlaced CCE encode on another, I'm going to compare them on the big screen probably tomorrow and find out what is the best way of encoding interlaced episodic material !!! Who knows, I might end up writing a small guide for the community !

Anyhow, thanks again for the help, I will tell you about my progress and if anybody else has tips for episodic encodes, that would be appreciated as well.

Regards. Pridebowl.

[TheSeeker]

I have just been looking more into the Telecide() filter. And I think you may be confused as to its intended usage. Its not, I repeat NOT, intended as a general purpose deinterlacer. Telecide() is a 3:2 pulldown remover. Here is an explanation of it from the Decomb readme:

Telecide(): Recovers progressive frames (by finding and aligning matching fields) but does not remove resulting duplicates and does not change the frame rate or frame count. Do not use Telecide on streams that do not contain telecined progressive frames, such as pure interlaced video. Note that a stream of PAL progressive frames that are shifted by one field will exhibit combing and can be considered a telecined stream for purposes of recovery; Telecide will easily handle this situation.

But since your source is a tv show that was probably originally shot at 29.97 fps it doesn't have any dropped frames to recover.

Scharfis_brain or someone like that would know for sure. But I think your usage of the telecide functions is incorrect. Im just thinking that there is no reason that your That 70's Show dvd's should contain telecined progessive frames. Since it was most likely shot and filmed 29.97 pure interlaced.

Can anyone back me up here? Or am I mistaken?

[Pridebowl]

You are probably right... But the result is no longer interlaced. One thing I have found watching the resulting m2v files is that every fifth frame is a copy of the fourth one like : 1-2-3-4-4-5-6-7-8-8-... and so on. Maybe that was a bad filter, but I tried other Deinterlace filters but they all gave me artifacts or blurried vertical smudge effect (TomsMoComp(1,5,1))

BTW, the Procoder 2.0 encode has finished after 750 minutes, so 12 hours and 30 minutes.

[TheSeeker]

You are probably right... But the result is no longer interlaced. One thing I have found watching the resulting m2v files is that every fifth frame is a copy of the fourth one like : 1-2-3-4-4-5-6-7-8-8-... and so on. Maybe that was a bad filter, but I tried other Deinterlace filters but they all gave me artifacts or blurried vertical smudge effect (TomsMoComp(1,5,1))

BTW, the Procoder 2.0 encode has finished after 750 minutes, so 12 hours and 30 minutes.

That whole every fifth frame thing is how telecide works. It drops every nth frame (n being 5 by default), and since the stream does not contain any telecined progressive material all it does is copies the 4th frame.

If I were you I would do some more research and verify that Telecide is the correct choise for you. Also, some other good filters I have had some good experiences with are SmoothDeinterlacer, and MipSmooth. You have to play around and tweak the arguments a little with those ones (although Im pretty sure Mipsmooth has settings presets like AnimeHQ and MovieHQ for example). Check those two out if your dead set on deinterlacing.

What filters did you use with Procoder 2.0. Because I think you might have liked the output without any filters at all. (or maybe with just a light Undot())

[kbello]

i got the same problem with lastest versions of dvdrb with Taken NTSC r1, in pc play fine but in TV got interlaced artifacts, very weird... i test all encoders...

[TheSeeker]

i got the same problem with lastest versions of dvdrb with Taken NTSC r1, in pc play fine but in TV got interlaced artifacts, very weird... i test all encoders...

Do the interlacing artifacts show up when you watch the originals on your TV? It may be that either your tv or your dvd player have fairly low end deinterlacing chips. The problem is that unless you buy high end tv's or dvd players most manufacturers save money by skimping on the deinterlacing hardware.

I remember doing Taken a while back. I remember a few jaggies still being present on the backup copies, but I dont remember it being all that noticeable, unless your specifically looking for it.

And lets all remember folks, that due to the way many of these films were shot/master/authored. There is almost no way to get them completely free of artifacts. Unless you want to sacrifice some serious detail. Or unless you have professional grade hardware, and optimized filters, and hours upon hours to optimize a professional filter to work just for that one dvd. At some point you just have to say. Ok, this is good enough, Im not willing to put in days upon days of work just to get a single dvd backed up.

[lark]

That whole every fifth frame thing is how telecide works. It drops every nth frame (n being 5 by default), and since the stream does not contain any telecined progressive material all it does is copies the 4th frame.

just a minor correction. it's not telecide that drops the frames. it just finds out the matching fields from different frames to produce uninterlaced (progressive) frames.

you need to use decimate(n) n defaults to 5 to remove 1 of every 5 frames. decimate finds out 2 frames that are most similar and drops one of them. but please note that this changes the framerate and i don't think you want that to happen in your NTSC DVDs...

regards
t :)

[Pridebowl]

TheSeeker, I only used Undot() for the ProCoder encode.

I now realize that my tests with Telecide() are worthless with that last two posts...

I'm gonna check SmoothDeinterlacer and MipSmooth, I didn't know about those two !

Thanks.

[TheSeeker]

How did the procoder output look? Much better than the CCE output?

[Pridebowl]

Hi,

I just bought a LD LDA-531 to make sure that my tests are okay. Verified, Canopus Procoder 2.0 does a better job than CCE for interlaced material.

I used Undot() as Avisynth filter, not Deen() because it is spacial.

Thanks for the help.

[manono]

Hi-

But the result is no longer interlaced, One thing I have found watching the resulting m2v files is that every fifth frame is a copy of the fourth one like : 1-2-3-4-4-5-6-7-8-8-... and so on.

Your material isn't interlaced, at least not in the sense of being created with 30fps video cameras. It was created on 24fps Film, edited in video, and encoded interlaced 29.97fps. This is quite common with TV series DVDs. The common term for this is hard telecine. It can and should be IVTC'd. Unfortunately, DVD-RB doesn't support IVTC, and you'll get better results by doing the encoding yourself and then running pulldown afterwards. And as lark pointed out, Telecide realigns the fields, while Decimate removes the duplicate frames. You'd need something like:

Telecide(Order=1,Guide=1).Decimate(5)

in your script to return it to 23.976fps. Don't try that in DVD-RB, though.

The reason it looks bad on your TV set is because (almost certainly) your progressive scan DVD player is a flag reader, interprets it as being interlaced, and deinterlaces the whole thing. If you had a cadence reader, it could perform the IVTC on-the-fly, so-to-speak, to give you back nice clean frames. If you watch a lot of TV series DVDs and/or anime, it might be worth your while to invest in a better DVD player. As TheSeeker said:

It may be that either your tv or your dvd player have fairly low end deinterlacing chips.

My guess is that the original DVDs and your reencode of them play with kind of a "strobing" effect, as out of every 5 frames, 2 are interlaced and 3 are clean progressive frames. If you played the one you used Telecide on, then you'll get a slight pause every 5th frame, making for jerky playback, very annoying during certain kinds of movement, such as a slow panning shot.

Keeping it interlaced is fine for watching on regular interlaced TV sets (except for the bits wasted reencoding 29.97fps for DVDR, as opposed to reencoding 23.976fps), but can present real problems when watching on an HDTV or computer (all the software players have the same problems). Just try stepping through it frame by frame in PowerDVD or Media Player Classic. Don't just deinterlace it. That will completely wreck it. Either keep it interlaced, or (much better) IVTC it.

What's an LD LDA-531? Do you mean LG LDA-531? What chipset does it use?