[MEncoder-users] Flashvideo for YouTube using mencoder

[Nicolas George]

L'octidi 8 ventôse, an CCXVI, Ben a écrit :
> The keyframe will not be compressed harder or stronger, it
> deliberately holds more value to the codec and stream. The keyframe will
> be bigger in size than any other frame in the stream. Hence its quality
> will be higher.

This reasoning is completely false. The whole point of video encoding,
versus encoding a sequence of still image almost all the same, is to achieve
the same quality with a much lower bitrate, because each image takes profit
from the information present in the previous one.

In other words, if you want to compare image quality by comparing the amount
of information present in it, you must take into account all the required
information. For an I-frame, that is just the data of the I-frame itself.
But for a P-frame, that is the data of the P-frame itself, but also the data
of all previous frames until the last preceding I-frame.

I just did the test with some DVD content I have near at hand (downscaled to
320×240 for speed): PSNR is 54.094882 for both settings, correlation is
0.999940, but it requires 40% less data with P-frames than without.

>		  Which by the way is clear to test too; Leave out your
> bitrate requirement, encode a video with keyint=1 and encode the same
> video with keyint=5000, you'll see which one is best.

Well, if you leave out all bitrate requirements, then you will get lossless
quality, provided the codec is able to do it. Thus, the quality will be
precisely the same. The bitrate, on the other hand, will be much smaller if
you allow P-frames.

Regards,

-- 
  Nicolas George