Title :
Comparison of motion vector detection based on gradient method for video sequences including large motions
Author :
Imamura, Kousuke ; Takabayashi, Masanori ; Hashimoto, Hideo
Author_Institution :
Fac. of Eng., Kanazawa Univ., Japan
Abstract :
The full search method for the block matching algorithm is a typical one for motion vector detection. The full search method, however, is not suitable for motion vector detection of a sequence including large motions, because of its high computational complexity. The motion vector detection based on the gradient method has the ability to detect motion vectors more efficiently. The purpose of this paper is to compare and improve motion vector detection methods based on the gradient method, particularly for video sequences including large motions.
Keywords :
Newton method; computational complexity; gradient methods; image sequences; motion compensation; motion estimation; video coding; Gauss Newton method; block matching algorithm; computational complexity; conjugate gradient method; full search method; gradient method; large motions; motion compensated prediction; motion vector detection; steepest descent method; video sequences; Codecs; Digital TV; Equations; Gaussian processes; Gradient methods; Motion detection; Newton method; Redundancy; Search methods; TV broadcasting;
Conference_Titel :
Signal Processing, 2002 6th International Conference on
Print_ISBN :
0-7803-7488-6
DOI :
10.1109/ICOSP.2002.1181177