Implementation of a covariance-based principal component analysis algorithm with a CUDA-enabled graphics processing unit

Author

Zhang, Jian ; Lim, Kim Hwa

Author_Institution

Centre for Remote Imaging, Sensing & Process., Nat. Univ. of Singapore, Singapore, Singapore

fYear

2011

fDate

24-29 July 2011

Firstpage

1759

Lastpage

1762

Abstract

There are three major approaches of principle component analysis (PCA [1]): singular value decomposition (SVD [2]), covariance-matrix and iterative method (NIPALS). This paper implemented these methods for medium-sized hyperspectral images [3, 4, and 5] in NVIDIA CUDA and compared the performance between them and their CPU counterparts. It is found that the covariance-matrix approach has a great potential of reaching a real-time performance.

Keywords

computer graphic equipment; coprocessors; covariance matrices; image processing; iterative methods; principal component analysis; singular value decomposition; CUDA-enabled graphics processing unit; compute unified device architecture; covariance matrix; covariance-based principal component analysis; iterative method; medium-sized hyperspectral image; singular value decomposition; Algorithm design and analysis; Graphics processing unit; Principal component analysis; Real time systems; CUDA; GPU; PCA; covariance; hyperspectral;

fLanguage

English

Publisher

ieee

Conference_Titel

Geoscience and Remote Sensing Symposium (IGARSS), 2011 IEEE International

Conference_Location

Vancouver, BC

ISSN

2153-6996

Print_ISBN

978-1-4577-1003-2

Type

conf

DOI

10.1109/IGARSS.2011.6049460

Filename

6049460