[x264-devel] MSU MPEG4-AVC/H.264 codecs comparison

[Jason Garrett-Glaser]

2010/2/1 Dmitriy Vatolin <dmitriy at graphics.cs.msu.ru>:
> Dear Developers,
>
>  As usual Moscow State University Graphics & Multimedia Laboratory
>  starts next 6-th H.264 codecs comparison.

The attachment provides no settings at or all or explains how you
chose between "psy" and "no psy".  Does "no psy" mean --tune psnr,
with AQ off?

> 3. Also for the future questions for companies: What usecases are
>   the most interesting for you?

I think the 1-pass ABR test should be replaced by a much longer CBR
test.  ABR is nearly useless in practice, while CBR is a huge use-case
for broadcast and so forth.  It should also be a longer test with many
scenes, to avoid biasing heavily based on the initial quantizer choice
as in Akiyo.

Some potentially interesting tests:

1) Varying VBV size: how do encoders do with normal VBV sizes, as well
as insane VBV sizes (a few frames, or even single-frame)?

2) Of encoders with lookahead, how much worse do they get when you
force them into a zero-latency mode (no lookahead)? (only reasonable
if many encoders allow this to be done)

Dark Shikari