[MEncoder-users] Finding good parameters for lavc mpeg4 encoder (cmp, subcmp, rate control equation etc. )

mesecam at gmail.com mesecam at gmail.com
Wed Apr 1 10:46:11 CEST 2009


Hi all,

I'am struggling with MEncoder (rev 28311) in order to find a good rate  
control equation (by good, I mean, one that can beat Xvid psnr figures.)

As a starting base, i've used this rc equation:

vrc_eq=(tex+100000000*mcVar)^qComp

from Corey Hickey http://fatooh.org/ratecontrol/

I came up with this:

first pass:
--------------
-ocv lavc -lavcopts  
vcodec=mpeg4:precmp=0:cmp=0:subcmp=0:mbcmp=0:predia=-1:dia=-1:vmax_b_frames=2:
vb_strategy=2:vb_qfactor=1.5:vb_qoffset=0.75:preme=2:last_pred=2:vrc_eq=\(mcVar*mv*10^3+tex\)*isP\)^0.7

2nd pass:
--------------
vcodec=mpeg4:precmp=256:cmp=256:subcmp=256:mbcmp=256:predia=-1:dia=-1:vmax_b_frames=2:
vb_strategy=0:vb_qfactor=1.5:vb_qoffset=0.75:preme=2:last_pred=2:mbd=2:trell:v4mv:bidir_refine=4:vrc_eq=\(mcVar*mv*10^3+tex\)*isP\)^0.7

vrc_eq is empirical, but every time it gives better figures than the  
default one and around ~0.2 psnr point better than Xvid,
95% of the time (but it's a slower encoding process than Xvid )

Any suggestions to improve this or better formulas etc ?


More information about the MEncoder-users mailing list