Re: [rtcweb] MTI motion JPEG

Actually that may work. Toying around with mjpeg I found that, for 
instance, CIF video can be transmitted with somewhat usable quality at 
full framerate at slightly over one mbit/s. Usable quality at around 
half a mbit/s can be achieved by decreasing the framerate. Perceived 
picture quality can be increased by postprocessing (deblocking, 
deringing) or the decoding technique described in "William B. 
Pennebaker, Joan L. Mitchell - JPEG: Still Image Data Compression 
Standard, 1993". I would also assume that support for the arithmetic 
encoder could be mandated by now, increasing coding efficiency.

So this makes for a lousy video codec, but at least it's fast, 
"everywhere", presumably free of additional IPR risk, and flexible 
regarding frame size. I can also be used to occasionally push 
high-quality pictures (think slide shows), a use case H.261 included a 
hack for (4CIF) but would be better handled by MJPEG anyways.

Maik

Am 30.11.2013 01:49, schrieb Ralph Giles:
> I'd like propose motion JPEG as the manditory-to-implement video codec,
> per RFC 2435 or similar.
>
> Yes, I'm serious. I didn't propose this earlier to avoid splitting the
> vote at the last meeting.
>
> Our goal here is to to avoid negotiation failure. We don't have
> concensus for either h.264 or VP8, but no one is happy with the
> performance of the older formats we've been discussing.
>
> Since we can't have acceptable performance in our fallback we should
> instead optimize for minimum complexity. We all have to ship this and
> hope it's never used in practice.
>
> While not as trivial as G.711 audio, mjpeg+rtp is simpler to implement
> than the more recent obsolete codecs we've been discussing. JPEG decode
> is already universally supported for still images. As the simplest
> compressed format it has advantages for testing and legacy
> interoperation, just like G.711. It works fine over a LAN and for the
> WAN...well, you can have a couple of blurry frames a minute if you like
> that better than nothing at all.
>
>   -r
> _______________________________________________
> rtcweb mailing list
> rtcweb@ietf.org
> https://www.ietf.org/mailman/listinfo/rtcweb
>