Four-step genetic search for block motion estimation

Author

So, Man F. ; Wu, Angus

Author_Institution

Electron. Design Autom. Center, City Univ. of Hong Kong, Hong kong

Volume

3

fYear

1998

fDate

12-15 May 1998

Firstpage

1393

Abstract

Genetic algorithms (GA) are well known for searching global maxima and minima. In general, number of search points required by GA for searching global extreme is much lower than the exhausted search. GA have been applied to the block matching algorithm (BMA) and demonstrate positively its capability in the BMA. The mean square error (MSE) performance of GA based BMA is close to full-search (FS). However, the disadvantage of GA is the computational requirement for practical use. A four-step genetic search algorithm is proposed for the BMA. The proposed method takes advantage of GA and 4SS. The simulation result shows the proposed method has a similar performance to full-search (FS) in terms of the MSE. In addition, the number of search points required by the proposed algorithm is approximately equal to 14% of the FS and is close to three-step search (3SS). The speed up ratio between the proposed algorithm and FS is 5.6 times

Keywords

data compression; genetic algorithms; image coding; image matching; image sequences; motion estimation; search problems; video coding; MSE performance; block matching algorithm; block motion estimation; exhausted search; four-step genetic search algorithm; full-search; genetic algorithms; global maxima; global minima; image sequence compression; mean square error; simulation result; speed up ratio; three-step search; video compression standards; Computational efficiency; Design engineering; Distortion measurement; Electronic design automation and methodology; Genetic algorithms; Genetic engineering; Image coding; Motion estimation; Partial response channels; Transform coding;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech and Signal Processing, 1998. Proceedings of the 1998 IEEE International Conference on

Conference_Location

Seattle, WA

ISSN

1520-6149

Print_ISBN

0-7803-4428-6

Type

conf

DOI

10.1109/ICASSP.1998.681707

Filename

681707