[FFmpeg-user] Reencoding from H.264 to H.265

[Peter White]

04.09.2016 23:09, Sven C. Dack:
 > On 04/09/16 21:13, Peter White wrote:
 >> Hi,
 >>
 >> 04.09.2016 20:40, Sven C. Dack:
 >> ...
 >> > *** Please note that this is _NOT_ a comparison of encoders, but it is
 >> > meant to show the effect of the (down-)scaling filters on reencoding.
 >> > ***
 >>
 >> But what effects?
 >
 > I am not sure how I can help you. If you still haven't understood that
 > the filters are having a significant impact on reencoding than how else
 > can I show this?

So you showed that they have an effect. Acknowledged. My question had
a follow-up, though. Your conclusions are not correct. The best scaler
is the one producing the best visual quality *before* compression.
Lanczos and natural bicubic spline rank at the top in that regard. If
that also means one has to spend more bits on encoding the resulting
frames, that is utterly fine and justified. A higher bitrate can mean
higher quality, after all.
And if that means they get a *higher* quantizer, even better, since
that translates to a lower bitrate. Nothing more, nothing less. You,
OTOH assume that a higher quantizer also translates to lower quality,
which is not true, given all other quality related settings stay the
same. See your own sample files. Lanczos (the S/W version) results in
significantly lower bitrate, yet is vastly superior quality-wise
compared to nearest neighbor.

 > You see, I could simply tell people not to do reencoding. It would save
 > everyone some time, wouldn't it?

Then why don't you? ;)

 > But this kind of attitude is often seen as disrespectful and patronizing.

It doesn't have to. Look at Nicolas' reply. I was in the process of
writing one myself, when I noticed he beat me to the punch. His reply
is even better than mine could have ever been, so I did not bother
finishing.
And when one expressly asks for "expert opinions", how is a reply from
such individual ever not going to sound "patronizing" to some extent,
if it must be "don't do it, for this and that reason"? It is a
reasoned expert response. If one cannot live with that, why ask in the
first place?

 > People will continue to transcode videos no matter what.

And that's fine, if they do know what they are doing and have made an
informed decision. Or, they should at least know that they are doing
so against the express advice of the experts they asked.

 > So instead of telling people what they can and cannot do do I permit
 > any use of it and try to give people as much information as they
 > need for drawing their own conclusions and for making their own
 > decisions.

And I cannot let stand unanswered, what you said in your replies, so
the same people know that there is a difference of opinion. Plus, you
did not really answer the OPs question. Nicolas did. You brought
scaling into the discussion when nobody asked about it.
And the OP had done their homework already, saying they had tried
themselves and noticed no significant reduction in bitrate, when
reencoding to H.265 from H.264 *and* keeping roughly the same
perceptual quality.


Peter