作者: Jiefu Zhai , Keman Yu , Jiang Li , Shipeng Li
DOI: 10.1109/ISCAS.2005.1465738
关键词: Motion compensation 、 Artificial intelligence 、 Motion vector 、 Frame rate 、 Interpolation 、 Computer vision 、 Quarter-pixel motion 、 Block-matching algorithm 、 Motion interpolation 、 Computer science 、 Motion estimation
摘要: In low bit-rate video communication, temporal subsampling is usually used due to limited available bandwidth. Motion compensated frame interpolation (MCFI) techniques are often employed in the decoder restore original rate and enhance quality. this paper, we propose a low-complexity high efficiency MCFI method. It first examines motion vectors embedded bit-stream, then carries out overlapped block bi-directional estimation on those blocks whose regarded as not accurate enough. Finally, it utilizes vector post-processing compensation generate interpolated frames further reduce blocking artifacts. Experimental results show that proposed algorithm outperforms other methods both PSNR visual performance, while its complexity also lower than methods.