[XviD-devel] GME and Block ME

[Christoph Lampert]

Hi,

I am in some kind of dilemma with my GME: Sometimes, it often needs block
motion vectors to calculate start parameters. But block ME should have
data from GME available as predictors, too. 
Currently, I "solve" this by using _two_ block searches, one pre-analysis
before GME and the ordinary after GME. I can't just use MEAnalysis (for
B-frame decision) as pre-analysis, because it's too coarse. I need at
least 16x16 blocks in halfpel resolution, so just ordinary ME, except for
4MV. With MEAnalysis this make 3 step of block ME... suboptimal! 

Is anyone interested in a small discussion about how to get rid of this
double/trippe work? Syskin perhaps? 

Also, I noticed that block search always checks 12 or more positions 
before terminating, because of the many predictors and refinements we
have. With this, we left the path of PMVfast and EPZS to use thresholds
to check as few positions as possible, thus gaining speed. 
Instead, we use thresholds, but nevertheless check almost everything. 
Maybe out SAD-routines are simply too fast, so it's not even noticed, but
we should discuss this, too, one day. 

gruel