[x264-devel] x264 predictors

[Nadir Raimondo]

Hi people,

I'm trying to integrate into the x264 library a custom motion estimation 
algorithm based on ancillary data provided by some sensors.
I found this function in me.c that seems to do all the work:

void x264_me_search_ref( x264_t *h, x264_me_t *m, int16_t (*mvc)[2], int 
i_mvc, int *p_halfpel_thresh )

This doom9 link http://forum.doom9.org/showthread.php?p=1076816 
describes how the function works but it is really outdated (2005) and I 
cannot understand how
mvc array works.

Futhermore into x264_me_t structure there is another prediction (mvp array)

Watching x264_mb_analyse_inter_p16x16 function in analise.c seems that 
there are two different predictions:
x264_mb_predict_mv_16x16 that calculate mvp array and 
x264_mb_predict_mv_ref16x16 that calculate mvc arrays

How mvp and mvc arrays work and how are they calculated?

Is the x264_me_t mvp predictor obtained as a median of MVs of blocks 
around the current block?
Why are there multiple values for mvc?

Thank you in advance for your help