[MPlayer-dev-eng] [LONG] PSNR comparison, new source and results

[Romain Dolbeau]

Jan Knutar wrote:

> I've never so far when I've encoded myself and done some limited PSNR 
> tests first, seen that qpel would improve things. Would it be possible 
> that you include qpel or not in further tests? I'm not convinced of its 
> usefulness yet :D

My first tests included qpel-or-not ; on the "CT,HD" scene it gave:

"[qpel vs. null] Average: from 43.9142 to 44.282 (0.3678, 0.8376%) [2304 samples] bestDiff: 1.4, worstDiff: 0.12 pCount: 2304, nCount: 0"
(samples number is actually twice what it should, everything is checked twice)

Basically, there was 1152 pair of option w or w/o qpel,
and 1152 of them (100% :-) showed an improvement using qpel,
varying between 0.12 and 1.4 point, and 0.3678 on average.

At least on that scene/that size/that bitrate/that everything,
qpel is a definite win.

>>a - don't use vlelim & vcelim, at least on that kind of scene at
>>     that kind of bitrate
> 
> Did it ever show a psnr gain in any of your tests? I'd be surprised if 
> it did.

No. On the Matrix scene (2304 samples so 1152 of each),
vlelim diminished the PSNR of 1151 (1 was unchanged) and
vcelim diminisged the PSNT of 1150 (2 were unchanged).

I won't use them again, I think :-)

-- 
Romain Dolbeau