Inter prediction Modes

[ali_umair21]

Hi,

I am confused about how do we select which inter-prediction mode we need to chose. I mean how do we decide which mode gives us the minimum number of bits to store/encode. For example a 16x16 mode may result in more sum of absolute error but requires only one unit of overhead (interm of MB packet fields) as compare to if we encode the same area with 4x4 macroblocks which may result in less number of SOAE but require 4 units of overhead.

Hope I made question clear.

Thanks

[LoRd_MuldeR]

I guess you cannot "know" beforehand. You have to try and see which one gives the best result/compromise for "error -vs- bit cost". This is called rate-distortion optimization (RDO).