DocumentCode :
314597
Title :
A parallel computing and neural network implementation of LBG image vector quantization
Author :
Jiang, J.
Author_Institution :
Loughborough Univ., UK
Volume :
1
fYear :
1997
fDate :
14-17 Jul 1997
Firstpage :
27
Abstract :
In this paper, the popular LBG vector quantization algorithm is implemented and redesigned into a competitive learning neural network. Based on sequential learning, a further multi-layer parallel neural network is presented to improve the data throughput and training length in which a group of vectors can be processed rather than one within each cycle. Experiments carried out support that an alternative solution to the under-utilization problem is provided and improved performance is achieved in comparison with the sequential competitive learning neural network
Keywords :
unsupervised learning; LBG image vector quantization; competitive learning neural network; data throughput; multi-layer parallel neural network; neural network implementation; parallel computing; performance; sequential competitive learning neural network; sequential learning; training length; under-utilization problem;
fLanguage :
English
Publisher :
iet
Conference_Titel :
Image Processing and Its Applications, 1997., Sixth International Conference on
Conference_Location :
Dublin
ISSN :
0537-9989
Print_ISBN :
0-85296-692-X
Type :
conf
DOI :
10.1049/cp:19970847
Filename :
614985
Link To Document :
بازگشت