[MEncoder-users] nr value recommendations

Matyas mplayer.list at sustik.com
Mon May 11 17:20:43 CEST 2009


Thanks for all the comments!!

I have tried hqdn3d=2:1:2 as well as nr=250 on the same 1 hour 1920:1080 HD
source, a Nature show (Eagles of Mull) broadcast on KLRU.  (Why these parms?
Just pulled them out of my hat, for lack of better way...)

I cannot tell the results apart.  The source appears to be quite good quality
apart from the obvious frame rate conversion the broadcaster did (25 -> 29.97).

The stats with hqdn3d=2:1:2
x264 [info]: slice I:957   Avg QP:26.20  size: 24938  PSNR Mean Y:40.58 U:47.19
V:48.09 Avg:41.88 Global:40.77
x264 [info]: slice P:56294 Avg QP:28.73  size:  8176  PSNR Mean Y:39.32 U:46.13
V:47.06 Avg:40.62 Global:39.27
x264 [info]: slice B:50410 Avg QP:31.14  size:  2420  PSNR Mean Y:37.87 U:45.25
V:46.48 Avg:39.24 Global:38.04
x264 [info]: consecutive B-frames: 49.0% 51.0%
x264 [info]: mb I  I16..4: 18.7% 73.0%  8.2%
x264 [info]: mb P  I16..4:  4.3%  8.9%  0.5%  P16..4: 56.8%  8.2%  4.9%  0.0%
 0.0%    skip:16.4%
x264 [info]: mb B  I16..4:  0.4%  1.0%  0.1%  B16..8: 30.0%  0.6%  1.0%
direct: 2.5%  skip:64.4%  L0:36.0% L1:59.7% BI: 4.3%
x264 [info]: 8x8 transform  intra:66.1%  inter:71.7%
x264 [info]: direct mvs  spatial:84.9%  temporal:15.1%
x264 [info]: ref P L0  68.1% 14.9%  8.7%  3.1%  2.6%  2.6%
x264 [info]: ref B L0  83.6%  9.3%  3.8%  2.0%  1.2%
x264 [info]: ref B L1  96.0%  4.0%
x264 [info]: SSIM Mean Y:0.9539878
x264 [info]: PSNR Mean Y:38.653 U:45.727 V:46.800 Avg:39.987 Global:38.659
kb/s:1349.77

with nr=250
x264 [info]: slice I:971   Avg QP:26.32  size: 25328  PSNR Mean Y:40.36
U:46.97 V:47.83 Avg:41.66 Global:40.59
x264 [info]: slice P:53146 Avg QP:28.97  size:  8565  PSNR Mean Y:38.89
U:45.83 V:46.73 Avg:40.20 Global:38.97
x264 [info]: slice B:53544 Avg QP:30.99  size:  2359  PSNR Mean Y:37.68
U:45.12 V:46.30 Avg:39.06 Global:37.91
x264 [info]: consecutive B-frames: 42.0% 58.0%
x264 [info]: mb I  I16..4: 18.0% 74.1%  7.9%
x264 [info]: mb P  I16..4:  4.1%  9.2%  0.5%  P16..4: 59.1%  8.7%  5.2%  0.0%
 0.0%    skip:13.1%
x264 [info]: mb B  I16..4:  0.4%  1.0%  0.1%  B16..8: 30.3%  0.5%  0.9%
direct: 2.5%  skip:64.3%  L0:36.2% L1:59.9% BI: 3.9%
x264 [info]: 8x8 transform  intra:67.5%  inter:71.6%
x264 [info]: direct mvs  spatial:85.3%  temporal:14.7%
x264 [info]: ref P L0  65.9% 15.7%  9.3%  3.5%  2.8%  2.8%
x264 [info]: ref B L0  82.8%  9.7%  4.1%  2.1%  1.3%
x264 [info]: ref B L1  95.6%  4.4%
x264 [info]: SSIM Mean Y:0.9491193
x264 [info]: PSNR Mean Y:38.302 U:45.487 V:46.528 Avg:39.650 Global:38.422
kb/s:1349.79

The bitrate was 1350, I cropped to 1904:1072 and scaled to 952:536 with
Lanczos, the parms were for high quality (subq=7, bframes=16 ...)

The source material maybe very good for compression evaluation (flying birds,
water splashing...)  If anyone wishes to compare small segments (and the
original recorded), let me know,  I am willing to reencode with various parms.

Thanks again for the commments.
Matyas
-
Every hardware eventually breaks.  Every software eventually works.


More information about the MEncoder-users mailing list