浏览了一下TM5的mpeg2编码器

1.TM5 uses orginal source for full search but  uses reconstructed pic for half per search, why?

2.when searching motion, only use Y component and igonore U, V ?

3.when deciding dct_type, use  field correlation, but cannot understand totally. And, what is the advantage of different dct_type?

BTW: MSSG reference software is written in so good manner!!!!!!!