Horribly incapable of encoding interlaced contents?

[aabxx]

I don't know if this is a fault with xvid in particular or ASP as a whole, but doing a quant 2 with interlaced encoding gives me bitrates that are... well... they almost seem out of spec considering they always go above 10 000 kbps!

It almost makes me wonder if that much bitrate is REALLY needed... maybe xvid, rather than ASP as a whole, does not know how to handle interlaced content efficiently and just throws a lot of bits at it?

Anyone have any insight?

Anyway, with the current implementation it's useless for interlaced encoding, considering how many benefits you get by just sticking to MPEG-2.

[unskinnyboy]

Interlaced content at quant 2? What kind of footage was it? What kind of matrix was used? XviD throws a lot of bits at it, because you are probably asking it to. 10,000 kbps isn't out of spec (I am not sure what the max. allowed bitrates are, but surely they depend on the Profile@Levels), but in this case, you are better off using MPEG-2 and not MPEG-4, at least because of compatibility issues. At 10,000+ kbps, you are not going to be able to decode the resultant MPEG-4 stream in anything but a high-end PC anyway. Or else raise the quant, go for a 2-pass with a LBR matrix, consider deinterlacing etc.

[aabxx]

It's vhs-captures... when deinterlaced (tdeint and nothing else) and encoded progressive, the bitrate is around 5000-6000 kbps at quant 2...

When I try to do it interlaced (no processing), xvid suddenly goes to like 11-12mbits all the time!

I've done this on many clips with lots of different settings (of course, comparing same settings with same settings as far as possible), and always, encoding interlaced produces wicked bitrates.

I understood interlacing would be less effective, but that it would require twice the bitrate is a little bit insane... so I concluded xvid is rather ineffecient with interlaced stuff. but I would like to have other opinions just to learn more :)

[unskinnyboy]

As per Manao's post here (http://forum.doom9.org/showthread.php?p=569180#post569180), XviD lacks field-based motion compensation and hence is less efficient than MPEG-2 in handling interlaced content (not sure if this has been fixed yet).

[foxyshadis]

It's not nearly that bad, though. aabxx, you did select interlaced encoding in the xvid config and the correct field order, right? I've never seen enormous increases, although it's usually 30%-50% higher depending on what I'm compressing.

That's also a pretty huge indictment for a complete lack of facts: What matrix? What resolution? What framerate? What xvid version? Do you compare to size against bobbing, instead of against deinterlacing? Did you actually run a q2 mpeg2 encode and compare sizes?

[aabxx]

Anyone know if divx or any other ASP codec handles interlaced material better (that is, more efficiently) than xvid? (I'm gonna test divx later on anyhow).

[aabxx]

Well, the bitrate with divx is same-ish... I thought divx had more advanced handling of interlaced contents and therefore would be more effective? Quality-wise they both look excellent, but after many picture comparisons I think divx is a very tiny bit sharper.

Some interesting findings:
- The encoding stats windows says xvid is not encoding b-frames (when it should be) in interlaced mode. Are b-frames disabled in this mode? The dvx stat window shows b-frames are being encoded. EDIT: Actually, xvid does seem to encode b-frames when in unstricted mode. But not on HomeTheatre-profile. But divx does encode b-frames with that same profile. Hmmm.
- I actually saved 20% bits with xvid when turning on adaptive quantization... that's quite extreme, although I do not yet know if it is just the clip, or the interlaced-mode that caused this.

Conclusion: With B-frames and adaptive quantization enabled, I managed to get the bitrate down with ~40% in interlaced-mode... now it's looking much more acceptable. Additionally going half-D1 from full dvd-resolution, the bitrate is down to about 4000 to 6000 kbps for interlaced encoding at q2. That is very acceptable indeed!

Looks like I will be able to use xvid for my current job after all! And that is a good thing, using xvid always gives you that "good" feeling, if you know what I mean... it's like a friend! :)

[kypec]

Hi guys,

yesterday I tried to encode my first piece:
source footage = sitcom Frasier DVD R2 720x576 PAL 25.0fps
duration of a single episode = 0:22:44

After reading many tutorials and documentation I managed to do the following:
ripped the VOB+IFO with DVD Decrypter (IFO mode, main english AC3 2/0 192kbps audio track only)
created .d2v project with DGIndex 1.4.8 with audio track demuxed
encoded with XviD 1.1.0 final with use of VirtualDubMod+AviSynth script (nothing but MPEG2Source, no resizing nor filtering whatsoever)

Here are my XviD settings:
two-pass AS@L5, custom QM=EQM ULR v3, B-VOPs max 2/1.50/1.00
I aimed for target filesize (using XviD's bitrate calculator) around 180 MB which means approx. 1033 kbps average bitrate

After putting the AVI file to my standalone DVD player I've noticed sort of jerkiness during playback as if the people's movements were discontinuous.
I realized at that very moment that the footage was most likely interlaced:(

My question is: do you recommend to encode it with XviD's interlaced option turned on or to use some avs deinterlace filter to get progressive frames?
And if I do the latter which filters do you suggest to try?

According to check's comments here (http://forum.doom9.org/showthread.php?p=856044#post856044)
and Mug Funky's opinion here (http://forum.doom9.org/showthread.php?p=805359#post805359)

I'd rather go for deinterlacing beforehand mainly because of more efficient bitrate distribution across the file as well as better compatibility with software players.

[check]

Hi guys,

yesterday I tried to encode my first piece:
source footage = sitcom Frasier DVD R2 720x576 PAL 25.0fps
duration of a single episode = 0:22:44

After reading many tutorials and documentation I managed to do the following:
ripped the VOB+IFO with DVD Decrypter (IFO mode, main english AC3 2/0 192kbps audio track only)
created .d2v project with DGIndex 1.4.8 with audio track demuxed
encoded with XviD 1.1.0 final with use of VirtualDubMod+AviSynth script (nothing but MPEG2Source, no resizing nor filtering whatsoever)

Here are my XviD settings:
two-pass AS@L5, custom QM=EQM ULR v3, B-VOPs max 2/1.50/1.00
I aimed for target filesize (using XviD's bitrate calculator) around 180 MB which means approx. 1033 kbps average bitrate

After putting the AVI file to my standalone DVD player I've noticed sort of jerkiness during playback as if the people's movements were discontinuous.
I realized at that very moment that the footage was most likely interlaced:(

My question is: do you recommend to encode it with XviD's interlaced option turned on or to use some avs deinterlace filter to get progressive frames?
And if I do the latter which filters do you suggest to try?

According to check's comments here (http://forum.doom9.org/showthread.php?p=856044#post856044)
and Mug Funky's opinion here (http://forum.doom9.org/showthread.php?p=805359#post805359)

I'd rather go for deinterlacing beforehand mainly because of more efficient bitrate distribution across the file as well as better compatibility with software players.

As I said in that thread - encoding as interlaced is a bad idea unless you know you want/need it. I also didn't mention that many divx standalones don't even support it, so you don't really want to use it at all.

Your video is PAL which means you will just need a straght deinterlace - I've always liked TDeint: http://forum.doom9.org/showthread.php?t=82264
There are a LOT of other choices though, taking the time to mess around with each will probably be rewarding. If you don't care about encoding time you can always try mvbob, but it's rather slow.

[aabxx]

If bitrate is an issue, go with progressive. If it's not, keep it interlaced. :sly:

As for deinterlacers, I'd second the recommendation for tdeint. It is clearly better than leakkerneldeint and tomsmocomp in my testing, and I don't think you could do much better. On interlaced TV leak and toms often do a good job, but on handycam-footage tdeint is much better from my experiences.