Methods of inter-view motion prediction for multi-view video coding and 3D video coding are disclosed. The motion parameters of each sub-PU in one current PU are predicted according to the motion parameters of a corresponding reference block in a coded picture in the reference view.