[MPlayer-users] PSNR tests: mbd=2, mv0, trell, and cbp (and...

[Tuukka Toivonen]

On Tue, 14 Oct 2003, rcooley wrote:

>> Oh well, that explains why I need to use so ridiculously high bitrate
>> (2000 kbit/s for 640x480, for unscaled 768x576 even that is too low).
>You can get much better results.

Nah... my quality preferences are just so high.

>First of all, hqdn3d works quite well (as has been said many times

I always use hqdn3d.

>The second thing I would recomend is blur.  I often use
>'unsharp' with low values.  Most TV signals have all sorts of random

I haven't tried that, I guess I will. But I wonder if it would be better to
just scale the video down even more than blur it.

>Finally,  using vhq (if you aren't) improves the quality far more than
>just hqdn3d or unsharp can.

I'm using simply the best options for encoding that libavcodec has :)
And this does include vhq.

>From your bitrates, it sounds like you have rather noisy video, and you

Not really, after hqdn3d.

>just might do better to use MPEG1 (which disregards very fine details)
>rather than MPEG4 (which uses extra bits to preserve the details of the
>analog artifacts :-) ).

Huh? Are you saying MPEG-1 gives better quality than MPEG-4?!