Log in

View Full Version : maybe a problem with ehq80


Castor-Troy
8th July 2005, 23:10
Hi all,

I would like to know if it's a bug with the ehq mode : the difference between ehq 80 and 85 is that he first one show "moving background".

It's an extract from Brotherhood of the Wolves, encode in 1 pass quality 80%, unconstrained quality mode, inloop cutoffquant @11 (default, autorv10), linear quality mode, ehq 80/85.

the script :

loadplugin("C:\Multimedia\Video\Creation\AviSynth 2.55\plugins\DGDecode.dll")
clip = Mpeg2Source("E:\loups\loups.d2v",idct = 2)
clip = Crop(clip,2,68,-2,-68)
clip = degrainmedian(clip,limity=2,limituv=4,mode=1)
clip = bicubicresize(clip,640,272,0,0.75)
clip = undot(clip)
clip = limiter(clip)
Return clip

Files :
ehq80 (http://thecastortroy.free.fr/ehq80-1.rmvb)
ehq85 (http://thecastortroy.free.fr/ehq85-1.rmvb)

edit : read on M.P.C., overlay mixer

Sirber
9th July 2005, 04:32
Both should be identical, except the filesize.

Castor-Troy
10th July 2005, 01:50
But it is not, if u looked at the samples :(

Sirber
10th July 2005, 18:35
I saw your "moving background". Seems like motion errors to me. Also, it has squares:

http://www.detritus.qc.ca/files/snapshot20050710133403.jpg

Maybe you should increase inloop quants.

karl_lillevold
10th July 2005, 20:38
I would suggest using rate control, and not constant quality. Due to the low contrast of the source, you end up with very low bitrate in these sections when you use constant quality. The bitrate is only a little over 400 kbps! I think with rate control, the QP would have gone lower, and you would have gotten better quality. And yes, there is a differnce between EHQ 80 and 85. In 80 there are some shortcuts to speed up the encoding that do not appear to work very well in this example. 85 is recommended, and what is used for 'high' complexity. 80 would be 'medium-high', and not very commonly used, thus not too well tested.

Castor-Troy
22nd July 2005, 16:19
I used constant quality at 80 because it quite match with xvid@q2.
Something strange : @ehq60, there is no moving backgroung, moreover, quality looks better than @ehq85.

numbers (oPSNR) go in the same way :
ehq60 : 35.6974
ehq80 : 35.6645
ehq85 : 35.6776

I tried with 2pass @ bigger bitrate (~740kb/s).
ehq60 : 47.1678
ehq80 : 47.2569
ehq85 : 47.5757
But it was weird : ehq60 looks better than ehq80 (but ehq85>ehq60)

With a 2pass @~1400+ kb/s (~q2 xivd), on a longer clip (with this one include) :
ehq60 : 44.6974
ehq80 : 44.8691
ehq85 : 45.0168
Same conclusion with last comparison (~740) : ehq85 is the best, but ehq60 looks better than ehq80. ehq80 is the one which offer moving background (I dislike this very much).

I agree that ehq85 is recommended, but it's weird that ehq60 dont give this kind of detail compare to ehq80.

edit : - first comparison is @quality 80%.
- what about : http://forum.doom9.org/showthread.php?t=84142&page=3&highlight=sagittaire post #59 ?

Sirber
22nd July 2005, 17:01
Doesn't affect 1-pass CQ encode :)