DVD2SVCD & Hybrid streams

[Pko]

What does exactly do DVD2SVCD now (last version, I mean) when it finds an hybrid (NTSC + FILM) stream? Takes into account the % of FILM and NTSC the stream has?

I have posted a thread about this in Avisynth forum:

http://rilanparty.com/vbb/showthread.php?s=&threadid=21904

[axia]

Well my experience with v106b4, whilst backing up a criterion collection ntsc film to pal was that a warning came up and it said something like ivct encoding will take much longer, the resulting svcd is superb, 4 pass cce, o.60 1/3

[Pko]

Yes, but that probably was a FILM stream, that is, the MPEG in the DVD probably had only the 24fps images of the original film, with a "3:2 pulldown" flag that instructs the player to construct a "fake" NTSC stream (24fps gets reduced to 23.976 and then separated into frames re-weaved as a 29.97fps that is NTSC-standard-compliant).

The process for FILM to NTSC is called "telecine", and so the inverse (and sometimes really complex) process is call "inverse telecine" or IVTC for short. That could take a lot of time, but since you only need to compress around 24/30 of the frames (~24 per second instead of ~30), you get in practice 25% more bitrate for each frame and so is like if a stream of 2000 kbps got converted to 2500 kbps!!! so is no wonder than it looks great.

The problem comes to some shows filmed for video or TV like Sci-Fi shows or anime, that have some parts filmed "traditionally", that is, at 24fps like a movie with actors or at 18fps for the animation in anime or cartoons, but some other parts are computer generated graphics. Since that are made in a computer, where the expensive part is the design, then they are rendered to be fully NTSC compliant (they are intended to be watched on NTSC TVs) an so are rendered at truly 29.97fps interlaced.

Then, the final episode (this is usually made for TV series) is composed and the finale stream is in some parts pure NTSC and in some parts FILM (or something more complex with anime or cartoons). The question is, how do you process that?

If you treat it like FILM, in the NTSC parts there will be a conversion from 29.97 interlaced to 23.976 progressive, and so the image will be blurry and the movement will be jumpy.

If you treat it like NTSC, you can deinterlace or you can treat as interlace; you will lose that "25% bitrate" advantage anyway, but the NTSC parts will look OK. If you deinterlace, you will lose quality and time, but it seems that compressing to an interlace stream gives some trouble to some people, that was my question, if that is true.

And I do not really know what DVD2SVCD really does when it has to deal with a hybrid stream...

[sandchar99]

I recently ran into a hybrid DVD.

I suggest you look at this thread:

see this thread (http://forum.doom9.org/showthread.php?s=&threadid=19712)

I had excellent results when I let DVD2AVI set up it's project. It stated that the film was hybrid.

I used Decomb and modified the AVS script:

Telecide()
Decimate(mode=1,threshold=50)

This yielded a 29.97 frame rate with no interlaced artifacts.

[Pko]

Originally posted by sandchar99

I had excellent results when I let DVD2AVI set up it's project. It stated that the film was hybrid.

I used Decomb and modified the AVS script:

Telecide()
Decimate(mode=1,threshold=50)

This yielded a 29.97 frame rate with no interlaced artifacts.

Well, I've done that but the results were not good enough... not bad but not well enough either.

What we are doing is to deinterlace, and to duplicate/blend frames in a "smart" way. In the FILM part, we are creating new frames to compress out of nowhere, and that causes some slightly jerky motion sometimes and and also less quality since we are compressing frames we do not need to.

In the NTSC part, we are deinterlacing a originaly interlaced material, and so we make the stream a little blurry and create some ghosting.

With my method (encoding as interlaced), in the FILM part we are NOT creating new frames, simply we are enconding some *fields* twice. We suffer quality loss compared to IVTC because we are compressing more information than needed, but do not introduce new (blurred/ghosted) frames nor jerky motion if we duplicate/blend the wrong frame, since we are not duplicating nor blending anything (except the fields the original stream instruct us to). The NTSC part is encoded exactly as it was in the original, we preserve the interlacing information, and so we cannot do better.

Also, the stream take 40 - 50% less time to encode since deinterlacing/IVTCing is a complex process.

So, my question remains: why do we ever deinterlace interlaced/hybrid material for SVCD? is just a wrong practice or there is something more I am not aware of? I hope DVD2SVCD author will give us his opinion, perhaps encoding as interlace is something good to introduce as default in newer versions, for interlaced or NTSC-over-10% hybrid streams...