TY - GEN
T1 - Fast motion estimation algorithm based on edge block detection and motion vector information
AU - Liu, Qin
AU - Chen, Zhenxing
AU - Goto, Satoshi
AU - Ikenaga, Takeshi
PY - 2007/1/1
Y1 - 2007/1/1
N2 - Compared with previous standards, H.264/AVC adopts variable block size motion estimation (VBSME) and multiple reference frames (MRF) to improve the video quality. So full search motion estimation algorithm (FS) has to consume massive computation, to calculate every search point in the search window for every block type with multiple reference frame. Some mathematical analyses on edge block are presented in this paper. Moreover in most cases, search range for FS is fixed and large, which is not necessary for most cases. An adaptive search range scheme is provided in this paper. In summary, two fast VBSME algorithms are proposed in this paper: 1) subsampling method based on the block edge information is applied 2) adaptive search range based on motion vector information is utilized. Experiments demonstrate that combing our proposed algorithms with UMHS almost saves 92%-98% Integer Motion Estimation (IME) time compared with exhaustive search with only less than 0.5% bitrate increase or 0.04 db PSNR decline.
AB - Compared with previous standards, H.264/AVC adopts variable block size motion estimation (VBSME) and multiple reference frames (MRF) to improve the video quality. So full search motion estimation algorithm (FS) has to consume massive computation, to calculate every search point in the search window for every block type with multiple reference frame. Some mathematical analyses on edge block are presented in this paper. Moreover in most cases, search range for FS is fixed and large, which is not necessary for most cases. An adaptive search range scheme is provided in this paper. In summary, two fast VBSME algorithms are proposed in this paper: 1) subsampling method based on the block edge information is applied 2) adaptive search range based on motion vector information is utilized. Experiments demonstrate that combing our proposed algorithms with UMHS almost saves 92%-98% Integer Motion Estimation (IME) time compared with exhaustive search with only less than 0.5% bitrate increase or 0.04 db PSNR decline.
KW - Aliasing
KW - Motion estimation
KW - Multiple reference frame
KW - Search range
KW - Subsampling
UR - http://www.scopus.com/inward/record.url?scp=43749095398&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=43749095398&partnerID=8YFLogxK
U2 - 10.1109/ISPACS.2007.4445956
DO - 10.1109/ISPACS.2007.4445956
M3 - Conference contribution
AN - SCOPUS:43749095398
SN - 9781424414475
T3 - 2007 International Symposium on Intelligent Signal Processing and Communications Systems, ISPACS 2007 - Proceedings
SP - 590
EP - 593
BT - 2007 International Symposium on Intelligent Signal Processing and Communications Systems, ISPACS 2007 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2007 International Symposium on Intelligent Signal Processing and Communications Systems, ISPACS 2007
Y2 - 28 November 2007 through 1 December 2007
ER -