Deformable motion model for Frame Rate Up-Conversion in video coding
Shen Songwei, Fan Liang, Luo Kun · 2016
In June 2015, VCEG released the HMKTA which added a lot of new coding tools to improve the coding efficiency and Frame Rate Up-Conversion (FRUC) mode is one of the merge mode of inter prediction. FRUC adopts bilateral matching and template matching to derive motion information. Because FRUC need to further refine the sub-block's motion information after deriving the motion vector for current prediction unit (PU), it will involves huge computational complexity. This paper presents the third mode for FRUC called Deformable mode. The new mode uses motion information of the adjacent block to the four corners of the current PU to derive the motion information of every minimum partition (MP) in current coding unit (CU). The experimental results show the proposed algorithm can reduce the coding time of FRUC, and it can reduce the encoding time by 13.1% in a best case scenario when compared with the regular algorithm.