Re: [rtcweb] Gaelle Martin-Cocher's and Andrew Allen's choices (Was: Straw Poll on Video Codec Alternatives)

[Maik Merten <[e-mail]>]

Am 09.01.2014 20:37, schrieb David Singer:
> H.261 isn’t implemented, and I think few will even if mandated. Even when given a lot of bits I seem to recall it wasn’t great.  It has size and other odd limitations.

Just for the sake of completeness: Provided one can live with the 
resolution limitations, H.261 should and probably will always deliver 
better results than MJPEG for a given bitrate. So if one recalls H.261 
not being great even with "lots of bits", MJPEG will most likely be 
(much) worse. The ultimate image quality limit for H.261 is the smallest 
available quantizer, which allows for virtually artifact-free visual 
reproduction. Enclosed is screenshot of the p64 H.261 encoder using the 
smallest available quantizer on frame 42 of the sign_irene_cif sequence.

This is merely to show that H.261 is not by design locked to a picture 
quality of "always bad" (you did not claim that, but one could misread). 
In practice I fully expect H.261 to deliver workable quality in 
scenarios where MJPEG just cannot deliver video content.

This of course doesn't mean that MJPEG already being "everywhere" is not 
a very nice aspect or that the H.261 resolution restrictions are not 
unfortunate.


Maik