Any free MVC encoder yet?

[rica]

Hi.

Any free Multiview Video Coding (MVC) encoder is available yet?

[Blue_MiSfit]

Nope, not yet.

[Kurtnoise]

http://research.nokia.com/page/4988

http://wftp3.itu.int/av-arch/jvt-sit.../JVT-AD207.zip

[rica]

Great, thank you kurt.

[Blue_MiSfit]

Oh! Well, consider my comment retracted, then!

Derek

[popper]

given the new Media Gateway Devices BCM7422 SOC from Broadcom, id expect to see these SVC/MVC STB's start appearing by this time next year perhaps 18 months from now so an open SVC/MVC Encoder that doesn't suck might be a good thing sometime sooner.

http://www.multichannel.com/article/...l_Res_3DTV.php

"
System-on-Chip Aimed at Media Gateway Devices

By Todd Spangler -- Multichannel News, 12/14/2010 3:40:24 PM

Broadcom wants to take cable subscribers deeper into the third dimension, with a supercharged set-top system-on-chip that has more than twice the horsepower of previous generations to deliver full-resolution 3DTV.

The BCM7422 -- designed for media gateway devices that stream IP video in the home -- provides a 1.3-Gigahertz central processor, along with H.264/MPEG codecs that enable 1080p high-definition video and full-resolution HD 3DTV. The SoC also features a 1-gigapixel-per-second 3D graphics processing unit for advanced 3D graphics acceleration, as well as home networking connectivity including MoCA and DLNA support.
...
The new Broadcom chip supports both Scalable Video Coding (SVC) and Multiview Video Coding (MVC) standards. MVC, the spec used in 3D Blu-ray Disc players, uses two separate full HD streams, whereas SVC uses an "enhancement" stream for the second-eye image that can reduce bandwidth by 40% to deliver full 3D video...."

[Dark Shikari]

Close enough?

[shon3i]

Originally Posted by Dark Shikari

Close enough?

What is potential usage of this?

[Dark Shikari]

Originally Posted by shon3i

What is potential usage of this?

Encoding 3D video. What else?

[LoRd_MuldeR]

Originally Posted by shon3i

What is potential usage of this?

I think for the user, this is the important part:
http://git.videolan.org/?p=x264.git;...b88dfa7edbccb6

[Disabled]

So now x264 is a full fledged MVC encoder? Is MVC really only adding those Flags (and reordering the two streams accordingly)?

*edit* This post clarified it. This is "only" a patch to create 3D flagged AVC and not MVC.

[Dark Shikari]

Originally Posted by Disabled

So now x264 is a full fledged MVC encoder? Is MVC really only adding those Flags (and reordering the two streams accordingly)?

*edit* This post clarified it. This is "only" a patch to create 3D flagged AVC and not MVC.

But they both serve the same purpose. Additionally, it's probably possible to create the referencing structure in which each view references the other (using frame interleaving) within AVC.

[Stephen R. Savage]

Dark Shikari: Is one stereoscopic packing mode more effective compression-wise than the others? Column interlacing and checkboard look pretty awful to me in this aspect, but I'm not sure how top/bottom compares with left/right.

[Dark Shikari]

Originally Posted by Stephen R. Savage

Dark Shikari: Is one stereoscopic packing mode more effective compression-wise than the others? Column interlacing and checkboard look pretty awful to me in this aspect, but I'm not sure how top/bottom compares with left/right.

All pixel-based packing modes are pants-on-head retarded.

Top/bottom is better than left/right for parallelism. left/right should probably never be used.

Frame interleaving may be better for compression, especially once x264 gets awareness of the frame interleaving occurring. I will probably be working on this after the end of this semester.

[shon3i]

@Dark Shikari, tell me how is possible to use two inputs? Left eye and Right eye, and pack that into one stream?

[Dark Shikari]

Originally Posted by shon3i

@Dark Shikari, tell me how is possible to use two inputs? Left eye and Right eye, and pack that into one stream?

Using any of the 6 methods listed in --frame-packing. All could be done with dead-trivial Avisynth scripts.

[rica]

http://www.pantarheon.org/AviSynth3DToolbox/

avisynth script > MVC encoding capable x264- inside MeGui...

[kieranrk]

Originally Posted by Dark Shikari

left/right should probably never be used.

If I recall correctly the argument for using left/right is a loss in horizontal resolution rather than vertical resolution with top/bottom (a la anamorphic).

[Dark Shikari]

Originally Posted by kieranrk

If I recall correctly the argument for using left/right is a loss in horizontal resolution rather than vertical resolution with top/bottom (a la anamorphic).

If you're cutting down resolution, that's different. I meant if you weren't cutting resolution.

[Blue_MiSfit]

The bits of test encoding I did last year was mostly left/right encoded, stuffed in a 1920x1080 frame, so each view got ~ 960x1080 pixels. I had assumed the same as kieranrk, that the loss in horizontal resolution was preferable to vertical resolution.

How would you encode two pictures side by side without cutting resolution? I'd guess Double width frames, i.e. 3840x1080 or 2560x720? Interesting!

Also, @eddman:

Say we get a good MVC encoder, but shouldn't our source come from two cameras? Weren't avatar's live-action scenes filmed that way?

Absolutely. That's the only proper way to do 3D for live action.

Derek