DocumentCode :
1225621
Title :
PPHPS: a parabolic prediction-based, fast half-pixel search algorithm for very low bit-rate moving-picture coding
Author :
Du, Cheng ; He, Yun ; Zheng, Junli
Author_Institution :
Dept. of Electron. Eng., Tsinghua Univ., Beijing, China
Volume :
13
Issue :
6
fYear :
2003
fDate :
6/1/2003 12:00:00 AM
Firstpage :
514
Lastpage :
518
Abstract :
In hybrid moving-picture coding schemes, motion estimation is conducted in two stages, one at integer pixel accuracy and the other at subpixel accuracy. In very low bit-rate coding, some fast motion-search algorithms are able to decrease their search points to about ten pixels at the integer pixel search stage. There remains, however, a need to study fast computing for the subpixel motion estimation. A novel, fast-search algorithm using parabolic prediction is proposed in this paper particularly for the half-pixel motion vector estimation. It needs to check only three half-pixels rather than the eight pixels, which are used in the full half-pixel motion search scheme (FHPS). Experimental results show that this algorithm preserves the image quality and has little influence on the bit-rate when compared to that of the referenced FHPS strategy.
Keywords :
data compression; image coding; motion estimation; prediction theory; search problems; PPHPS; bit-rate; computational complexity; fast half-pixel search algorithm; fast motion-search algorithms; half-pixel motion search scheme; half-pixel motion vector estimation; hybrid moving-picture coding; image quality; integer pixel accuracy; integer pixel search; motion estimation; moving-picture coding; parabolic prediction-based search algorithm; subpixel accuracy; subpixel motion estimation; very low bit-rate coding; Acceleration; Digital communication; Frequency selective surfaces; Helium; Image quality; Interpolation; MPEG 4 Standard; Motion estimation; Prediction algorithms; Predictive models;
fLanguage :
English
Journal_Title :
Circuits and Systems for Video Technology, IEEE Transactions on
Publisher :
ieee
ISSN :
1051-8215
Type :
jour
DOI :
10.1109/TCSVT.2003.813416
Filename :
1207408
Link To Document :
بازگشت