Core Concepts
Efficiently implementing PCA on high-performance devices for hyperspectral image processing.
Abstract
The article discusses the implementation of PCA on high-performance devices for hyperspectral image processing.
It covers the importance of dimensionality reduction in hyperspectral imaging.
Details the steps involved in PCA implementation, including image preprocessing, covariance computation, eigenvector decomposition, and projection.
Utilizes CUDA, cuBLAS, and Thrust libraries for parallel computing on GPUs.
Highlights the comparison of different high-performance computing platforms for PCA implementation.
Stats
"Received: 20 April 2018; Accepted: 30 May 2018; Published: 1 June 2018"
"Hyperspectral imaging systems provide detailed information of the Earth's surface."
"PCA is a popular method for spectrally compacting high-dimensional datasets."
"The Jacobi method is used for eigenvector decomposition in PCA."
"GPUs and manycores are popular high-performance computing platforms for hyperspectral image processing."
Quotes
"Dimensionality reduction represents a critical preprocessing step in order to increase the efficiency and the performance of many hyperspectral imaging algorithms."
"PCA is based on projections techniques seeking for the best subspace for representing the collected hyperspectral data."
"Different research groups have published works that deal with the implementation of the PCA algorithm onto different high-performance computer architectures."