Log in

View Full Version : Observations with 2 very different encodes


Blue_MiSfit
1st January 2004, 04:01
Hello,

I have been playing with Selam recently, and I have noticed a few things. I did several 2-pass encodes including the following:

Lock, Stock, and Two Smoking Barrels:
Simple AviSynth script - nothing special besides UnDot.

Full DVD Resolution
1.2 GB Target Size
MPEG Quant
Motion Search 6
VHQ 4
Qpel, Adaptive Quant, Trellis, BVOPS @ 1, 1.5, 1
I-Frame interval: 300

Comments: Absolutely beautiful. This is as close to perceptably lossless that I have ever seen. Perfect high-motion, no macroblocking, sharp textures... Perfect.


Das Boot:

Same settings, but with resolution 608x344, and target size 1.4 GB
(700mb per DVD side)

Comments: This movie is very noisy, and overall displays much less contrast than the other film. As such, XviD has a lot of trouble with it (see below).

1) the codec does EXTREMELY well in high motion scenes with a lot of contrast. (Explosions on gray background aka the end of Das Boot)

2) on the flipside - the codec has a lot of trouble with barely moving scenes especially dark scenes with little or no contrast. These scenes get all kinds of "swimming" artifacts in my encodes.

That has been the one major problem I have been having with my encodes. If anyone has suggestions as to how I could increase my encoding quality please let me know!

Thanks!
~misfit

Koepi
1st January 2004, 10:36
Give other codecs a try, you'll see the same or worse effects there.

It's a problem with minimal contrast and general dark areas, which all codecs i know have. Use LumaFilter() in your avs-script and see if it helps a little.

Regards
Koepi

Blue_MiSfit
2nd January 2004, 01:55
thanks, will do!

~misfit

Teegedeck
3rd January 2004, 16:36
And maybe it would be a good idea to give adaptive quantization a miss on this one - as dark scenes apparantly are too highly-quantized anyway. I am not hopeful that it helps though. AQ works pretty well by no, so it probably wasn't used on the frames in question.