25fps BD with MBAFF to 23.976

[Anonymous344]

I've a 25fps BD encoded with MBAFF. Here's a sample that I trimmed out with DGAVCDecDI. Eac3to seems to be able to slow the video down to 23.976fps successfully. The resulting file still registers as MBAFF but appears to play back fine on a media player. What do you guys think? Are there any potential problems associated with slowing down an MBAFF encoded stream in this way?

[poisondeathray]

I don't think there will be any problems on a software media player, but some hardware devices might have problems, and re-authoring with blu-ray software will most likely reject the streams

This is 1080p25 content. The reason why it's prepared like this is 1920x1080p25 native is not in the blu-ray specs, and incompatible on standalone units (although some may play it), so they encode it MBAFF. It's analgous to PAL DVD's where 25p content is in a 50i stream.

[Anonymous344]

Ah, I see! Thanks for the information and explanation. I did wonder whether the disc could be re-authored. I've never seen a BD like this before, so I found it interesting. I've only really come across MBAFF with HDTV.

[poisondeathray]

MBAFF is just a more efficient method of encoding ; the encoder can decide whether to encode as field or frame

"PAFF" is the other method and more analgous to what we would think as "interlaced"

But the MBAFF flag in seq parameter heading will probably cause authoring software to reject that stream, unless it was 25fps or 29.97fps . If you re-encoded it as progressive, you could author it as 23.976 or 24.0p

[Anonymous344]

MBAFF is just a more efficient method of encoding ; the encoder can decide whether to encode as field or frame

Yeah, I understand that, but I've never understood what the ultimate advantage to doing this is when it's a movie that's being encoded.

Thanks for explaining about the MBAFF flag. I wonder if the flag could be removed to fool the authoring software. Obviously the result would probably cause some players problems.

[poisondeathray]

MBAFF is just a more efficient method of encoding ; the encoder can decide whether to encode as field or frame

Yeah, I understand that, but I've never understood what the ultimate advantage to doing this is when it's a movie that's being encoded.

Yes, in this case, the only reason is a "hack" to comply with blu-ray specs. But for interlaced content, it's signficantly more efficient (so better quality at a given bitrate)

I'm assuming this was a film piece, that makes you wonder why it wasn't released as 24p ?

I wonder if the flag could be removed to fool the authoring software. Obviously the result would probably cause some players problems.

I was thinking of that too, but I think it's more critical data, not like AR information or VUI information

roozhou has a modified ffmpeg tool that can change some things, but I don't think MBAFF is one of them

http://forum.doom9.org/showthread.php?t=152419

I would ask in the h264/avc forum at doom9, people there would know if this is something that could be done

[Anonymous344]

Yes, in this case, the only reason is a "hack" to comply with blu-ray specs. But for interlaced content, it's signficantly more efficient (so better quality at a given bitrate)

I see! So videos that have progressive and interlaced content could benefit from MBAFF. That sounds logical.

Thanks for the other suggestions. I was mostly curious about what could be done with this sort of stream, because it's the first time I have ever seen it on a BD, but I might follow the matter up at some point.

[poisondeathray]

Yes, in this case, the only reason is a "hack" to comply with blu-ray specs. But for interlaced content, it's signficantly more efficient (so better quality at a given bitrate)

I see! So videos that have progressive and interlaced content could benefit from MBAFF. That sounds logical.

Actually PAFF can do this too, at least theoretically. The "A" stands for adaptive. PAFF is "picture adaptive field frame", so it can adapt. But I've never seen any PAFF implementation that is actually adaptive. The difference for MBAFF is that it decides on a macroblock basis, not entire frame basis, so it's even more efficient

EDIT: actually there are adaptive examples of PAFF. I think BBC and some Italian channels used to use it for football (soccer) matches - but I think they have switched over to MBAFF

[Anonymous344]

Actually PAFF can do this too, at least theoretically. The "A" stands for adaptive. PAFF is "picture adaptive field frame", so it can adapt. But I've never seen any PAFF implementation that is actually adaptive. The difference for MBAFF is that it decides on a macroblock basis, not entire frame basis, so it's even more efficient

Lol! So it turns out that I understand why PAFF is used, but not why every aspect of MBAFF is used. I understand why, if a source is a mix of progressive and interlaced, one might want to encode some frames as fields and some as frames, but under what circumstances would one want to encode some macroblocks as interlaced but others as progressive. I can't understand that at all.

EDIT: actually there are adaptive examples of PAFF. I think BBC and some Italian channels used to use it for football (soccer) matches - but I think they have switched over to MBAFF

EDIT: By adaptive, in this context, do we mean the decision the encoder makes to encode individual frames as frames or fields? We're not talking adaptive quantization like HCEnc does, are we?

[poisondeathray]

EDIT: By adaptive, in this context, do we mean the decision the encoder makes to encode individual frames as frames or fields? We're not talking adaptive quantization like HCEnc does, are we?

yes . AQ is a different concept. It redistributes bitrate to flat less detailed areas, like gradients, skies, shadows. But HCEnc's implementation of AQ isn't that good .

Lol! So it turns out that I understand why PAFF is used, but not why every aspect of MBAFF is used. I understand why, if a source is a mix of progressive and interlaced, one might want to encode some frames as fields and some as frames, but under what circumstances would one want to encode some macroblocks as interlaced but others as progressive. I can't understand that at all.

It works out to be higher quality /more efficient - at least with metrics - visually is another story , because the metrics can deviate from human perception . But in terms of coding efficiency, progressive encoding is always better than interlaced. For example, with interlaced encoding, Motion estimation is worse, things like motion vectors get truncated




Actually if you look at the x264 development thread over at doom9, MBAFF was recently improved (the last 20-30 commits were all MBAFF related) and you can see comparison numbers there (before it was non adaptive, and many features were not implemented yet, even though it used MBAFF encoding)

[Anonymous344]

I'll have a look. Thanks again.

[poisondeathray]

I'll probably see it over at doom9, but if you ask and find out if you can somehow modify the MBAFF seq paramater heading, post it back here in case someone else has the same situation