[MPlayer-dev-eng] [RFC] preliminary x264 encoding help text

[Loren Merritt]

On Sat, 9 Apr 2005, Jeff Clagg wrote:

> And a special prize is deserved by anyone who can show a PSNR gain by 
> adjusting the deblocking parameters by -2 or more.

Customizable deblock filter was never about maximizing PSNR. 0(+-1) will 
always be best for PSNR. The reason they are customizable is that the 
standard allows it, it was easy to do, and we didn't know it would never 
help percieved quality until we tried it on lots of videos.

> And on a semi-related note, can anyone give a convincing justification for
> why the man page still states that increasing frameref is more effective
> on anime vs. live material? I fear this comment may have its origins in
> some poorly-considered and half-assed tests I did way back last fall.

I think it's still true. I know that values all the way up to 16 can 
greatly help anime. Based mostly on other peoples tests, I think values 
above 5 or so have negligible effect on most live-action content.

> b_adapt:
> With this option enabled, the encoder will use some simple heuristics
> to reduce the number of bframes used in scenes that might not benefit
> from them as much.

No, if a scene just didn't benefit as much from B-frames, I would still 
use them. The purpose of b_adapt is to reduce or disable B-frames in 
scenes where they would hurt compression.

> The speed penalty of adaptive bframes is currently rather modest, but 
> so is the potential quality gain. It usually doesn't hurt, however.

With b_adapt, increasing max B-frames should never hurt compression. 
Without b_adapt, it might. If b_adapt ever reduces quality, it's due to 
lack of tuning in my heuristic, not anything inherent in the option.

> b_pyramid:
> You might as well enable this option if you're using >2 bframes; as the
> man page says, you get a little quality improvement with no speed cost.

While I haven't found any movies that are hurt by b_pyramid, I still 
consider it to be an experimental option. It probably works, but be 
warned.
The way b_direct works in conjunction with b_pyramid, while good for fast 
decoding, has some potentially harmful side-effects which may not always 
be counteracted by the improved compression from pyramid. I have a couple 
of ideas for workarounds, but haven't got around to implementing them yet.

--Loren Merritt