[MEncoder-users] Call for video encoding settings

Mon Aug 15 22:11:21 CEST 2005

Hi,

On 8/15/05, Guillaume POIRIER <poirierg at gmail.com> wrote:
> Hi there,
> 
> I've had a little bit of time to test those settings and run a full
> benchmark of those settings on a 2000 frames sample.
> 
> I ran a constant quantiser encode to test the relative compressibility
> gain by each of the encode settings.
> 
> I guess those results would need a bit of "re-normalization" given
> that the PSNR are fairly different, which makes a good comparison more
> difficult (I guess a 2-pass encode with all those settings would help
> a lot, but that's all for today for me: ;-) )

Arf! Well, I did them anyway, adding Michael's Snow:

> Short analysis of the results: lavc is very good to compress VERY fast
> and well, XviD seems a bit better for fast High quality encode, and
> x264 kicks ass when encoding for very high quality.

In two pass mode, in High Quality settings x264 kicks ass and takes
names (PSNR-wise). Now, maybe a 'blind test" could give the "best open
source codec" crown to one of the ASP codecs, or maybe Snow, as H.264
is advantaged by the inloop filter (at least, that's my undertanding).

lavc is still the fastest and best codec when doing some fast/realtime
encode, and XviD is both fast and of good quality for normal encode
(note that 2-pass ratecontrol seems totally off by 10% in half of the
encode, when x264 has the best ratecontrol around).

Please discuss those results as I plan to add those "encoding
profiles" to MEncoder XML docs.

> On 7/20/05, Guillaume POIRIER <poirierg at gmail.com> wrote:
> [..]
> > * Very High Quality:
> > lacv: vcodec=mpeg4:mbd=2:mv0:trell:v4mv:cbp:last_pred=3:predia=2:dia=2:vmax_b_frames=2:vb_strategy=1:precmp=2:cmp=2:subcmp=2:preme=2:vme=5:naq:qns=2
> 
> 3fps, 1886.141 kbit/s, PSNR: Y:50.05, Cb:51.94, Cr:52.51, All:50.66
> 
> > XviD:
> > chroma_opt:vhq=4:bvhq=1:quant_type=mpeg
> 
> 12fps, 1748.798 kbit/s, Average PSNR y : 48.16 dB, u : 50.61 dB, v : 51.11 dB
> 
> > x264:
> > subq=6:4x4mv:8x8dct:me=3:frameref=5:bframes=3:b_pyramid:weight_b
> 
> 5fps, 1300.185 kbit/s, PSNR Mean Y:51.71 U:54.79 V:55.62 Avg:52.52 Global:50.80
> ---------------------------------------------------------
> >  * High quality:
> > lavc:
> > vcodec=mpeg4:mbd=2:trell:v4mv:last_pred=2:dia=-1:vmax_b_frames=2:vb_strategy=1:cmp=3:subcmp=3:precmp=0:vqcomp=0.6:turbo:psnr
> 
> 13fps, 1968.337 kbit/s, PSNR: Y:50.18, Cb:52.06, Cr:52.62, All:50.79
> 
> 
> > XviD:
> > vhq=2:bvhq=1:psnr:chroma_opt:quant_type=mpeg:psnr
> 
> 18fps, 1790.227 kbit/s, Average PSNR y : 48.14 dB, u : 50.64 dB, v : 51.14 dB
> 
> 
> > x264:
> > subq=5:4x4mv:8x8dct:frameref=2:bframes=3:b_pyramid:weight_b:psnr
> 
> 11fps, 1473.458 kbit/s, PSNR Mean Y:51.36 U:54.51 V:55.37 Avg:52.18 Global:50.59
> 
> 
> --------------------------------------------------------
> >  * Fast:
> > lavc:
> > vcodec=mpeg4:mbd=2:trell:v4mv:turbo:psnr
> 
> 34fps, 2043.167 kbit/s, PSNR: Y:49.89, Cb:52.04, Cr:52.57, All:50.56
> 
> 
> > XviD:
> > turbo:vhq=0:psnr
> 
> 28fps, 1898.784 kbit/s, Average PSNR y : 47.65 dB, u : 50.10 dB, v : 50.64 dB
> 
> 
> > x264:
> > subq=4:bframes=2:b_pyramid:weight_b:psnr
> 
> 15fps, 1554.404 kbit/s, PSNR Mean Y:51.37 U:54.37 V:55.21 Avg:52.14 Global:50.47
> 
> 
> -----------------------------------------------------------
> >   * Realtime:
> > lavc:
> > vcodec=mpeg4:mbd=2:turbo:psnr
> 
> 49fps, 2018.563 kbit/s, PSNR: Y:49.09, Cb:51.50, Cr:52.12, All:49.82
> 
> 
> > XviD:
> > turbo:nochroma_me:notrellis:max_bframes=0:vhq=0:psnr
> 
> 38fps, 2436.407 kbit/s, Average PSNR y : 48.88 dB, u : 50.71 dB, v : 51.20 dB

Snow, constant quality:
vqscale=2:vcodec=snow:cmp=1:subcmp=1:vqcomp=0.6:pred=1:psnr:vstrict=-2
8fps  941.722  kbit/s, PSNR: Y:45.70, Cb:47.96, Cr:48.61, All:46.40
vqscale=2:vcodec=snow:cmp=1:subcmp=1:vqcomp=0.6:pred=1:psnr:vstrict=-2:qpel
8fps  931.956  kbit/s, PSNR: Y:45.77, Cb:47.91, Cr:48.55, All:46.45

Two pass encode:
---------------------
All results are in decreasing level of the previously mentionned
quality settings (Very High Quality, High quality, Fast, Realtime)
The target bitrate was 900 kb/s

Lavc

6fps  909.944 kbit/s PSNR: Y:45.28, Cb:47.61, Cr:48.48, All:46.02
15fps 908.248 kbit/s PSNR: Y:45.24, Cb:47.58, Cr:48.40, All:45.97
42fps 906.681 kbit/s PSNR: Y:44.43, Cb:47.32, Cr:48.20, All:45.28
54fps 905.553 kbit/s PSNR: Y:43.92, Cb:47.03, Cr:47.91, All:44.81

XviD

14fps 894.029 kbit/s, Average PSNR y : 44.80 dB, u : 47.72 dB, v : 48.41 dB
20fps 895.003 kbit/s, Average PSNR y : 44.66 dB, u : 47.68 dB, v : 48.41 dB
29fps 996.446 kbit/s, Average PSNR y : 44.19 dB, u : 46.95 dB, v : 47.63 dB
41fps 993.631 kbit/s, Average PSNR y : 43.67 dB, u : 45.88 dB, v : 46.58 dB

x264
6fps  901.249 kbit/s, PSNR Mean Y:50.39 U:53.58 V:54.41 Avg:51.22 Global:48.84
13fps 901.408 kbit/s, PSNR Mean Y:49.53 U:52.89 V:53.75 Avg:50.39 Global:47.98
17fps 901.584 kbit/s, PSNR Mean Y:49.02 U:52.38 V:53.19 Avg:49.87 Global:47.36
(No realtime setting)

Two pass:
vqscale=2:vcodec=snow:cmp=1:subcmp=1:vqcomp=0.6:pred=1:psnr:vstrict=-2
9fps  898.809 kbit/s, PSNR: Y:45.48, Cb:47.79, Cr:48.45, All:46.19

vqscale=2:vcodec=snow:cmp=1:subcmp=1:vqcomp=0.6:pred=1:psnr:vstrict=-2:qpel
9fps  898.991 kbit/s, PSNR: Y:45.41, Cb:47.72, Cr:48.36, All:46.12

-- 
A legend is an old man with a cane known for
what he used to do. I'm still doing it.
  -- Miles Davis