[whatwg] <video> element proposal

[Maik Merten]

Bjoern Hoehrmann schrieb:
 > Flash supports two codecs, the more recent one is VP6, a successor of
> VP3; VP3 in turn is what Ogg Theora is based on. I would be surprised
> to learn that On2 gave the superior codec away for free while it sells
> the inferior one.

On2 VP6 is performing better than On2 VP3, on which Theora is based.

However, the original VP3 encoder IIRC had a few quality impacting bugs,
those are fixed in the Theora reference encoder. Plus the spec for
Theora has been extended to  overcome some limitations that were present
in the original VP3 (the reference encoder doesn't use these new
bitstream features yet, so there's room for improvement over what we see
today).

To put it into a nutshell: Theora is not just VP3 in an Ogg container.
It's a descendant to VP3 just like VP4, VP5, VP6...

I can't comment on how Theora compares to VP6, but I'm pretty sure
Theora outperforms H.263 which is said to be used at Google Video or
YouTube for compatibility reasons.

One example, using the football VQEG sequence (deinterlaced with a
linear blend deinterlacer, cropped to remove black borders, YV12,
uncompressed YUV4MPEG2 input):


ffmpeg2theora -x 352 -y 288 --aspect 4:3 -K 512 -v 0.5 football.yuv

http://datadump.da.funpic.de/football.ogg (474K)


mencoder -vf scale=352:288 -ovc lavc -lavcopts
vcodec=h263:vhq:vqscale=18:trell:keyint=512 -o football-h263.avi
football.yuv

http://datadump.da.funpic.de/football-h263.avi (478K)


As far as I know the ffmpeg H.263 encoder implementation used by
mencoder is of rather good quality and I tried to use the best settings
available. Anyway, this of course is no extensive benchmark at all (and
this list would be the wrong place anyway) but I think it gives a first
impression.

Personally I think the current draft by Ian Hickson ("UAs SHOULD support
Theora for video and Vorbis for audio, and SHOULD support the Ogg
container format") makes a lot of sense as it encourages the adoption of
one base format people can rely on and still doesn't hinder innovation
on the codec side.