TY - GEN
T1 - A Fast Multidimensional Data Fusion Algorithm for Hyperspectral Spatiotemporal Super-Resolution
AU - Chang, Pai Chuan
AU - Lin, Jhao Ting
AU - Lin, Chia Hsiang
AU - Tang, Po Wei
AU - Liu, Yangrui
N1 - Funding Information:
This study is supported partly by the Einstein Program (Young Scholar Fellowship Program) of Ministry of Science and Technology (MOST), Taiwan, under Grant MOST 110-2636-E-006-026; and partly by the Higher Education Sprout Project of Ministry of Education (MOE) to the Headquarters of University Advancement at National Cheng Kung University (NCKU). We thank National Center for High-performance Computing (NCHC) for providing computing and storage resources.
Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - In hyperspectral remote sensing, obtaining fine spatial-temporal and spatial-spectral resolution images are two critical fusion issues due to inherent optical sensor trade-offs. However, simultaneous realization of spatial, spectral, and temporal super-resolution is highly challenging. This paper formulates a new fusion framework incorporating all the three spatial/spectral/temporal dimensions to achieve hyperspectral spatiotemporal (HST) super-resolution. Additionally, unlike many fusion methods assuming availability of the spatial blurring matrix (SBM) in the forward model, we go a step further to blindly achieve HST super-resolution by automatically estimating the SBM. Subsequently, final results can be obtained through the fast iterative shrinkage-thresholding algorithm (FISTA) and the convex optimization-based coupled nonnegative matrix factorization (CO-CNMF) algorithm. We compare results of the proposed HST super-resolution method, termed GFCSR, with other extended spatiotemporal fusion frameworks designed for multispectral images, and, as it turns out, quantitative evaluations demonstrate the superiority of our algorithm.
AB - In hyperspectral remote sensing, obtaining fine spatial-temporal and spatial-spectral resolution images are two critical fusion issues due to inherent optical sensor trade-offs. However, simultaneous realization of spatial, spectral, and temporal super-resolution is highly challenging. This paper formulates a new fusion framework incorporating all the three spatial/spectral/temporal dimensions to achieve hyperspectral spatiotemporal (HST) super-resolution. Additionally, unlike many fusion methods assuming availability of the spatial blurring matrix (SBM) in the forward model, we go a step further to blindly achieve HST super-resolution by automatically estimating the SBM. Subsequently, final results can be obtained through the fast iterative shrinkage-thresholding algorithm (FISTA) and the convex optimization-based coupled nonnegative matrix factorization (CO-CNMF) algorithm. We compare results of the proposed HST super-resolution method, termed GFCSR, with other extended spatiotemporal fusion frameworks designed for multispectral images, and, as it turns out, quantitative evaluations demonstrate the superiority of our algorithm.
UR - http://www.scopus.com/inward/record.url?scp=85143121233&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85143121233&partnerID=8YFLogxK
U2 - 10.1109/WHISPERS56178.2022.9955073
DO - 10.1109/WHISPERS56178.2022.9955073
M3 - Conference contribution
AN - SCOPUS:85143121233
T3 - Workshop on Hyperspectral Image and Signal Processing, Evolution in Remote Sensing
BT - 2022 12th Workshop on Hyperspectral Imaging and Signal Processing
PB - IEEE Computer Society
T2 - 12th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing, WHISPERS 2022
Y2 - 13 September 2022 through 16 September 2022
ER -