Abstract
In order to reconstruct a high spatial and high spectral resolution image (H2SI), one of the most common methods is to fuse a hyperspectral image (HSI) with a corresponding multispectral image (MSI). To effectively obtain both the spectral correlation of bands in HSI and the spatial correlation of pixels in MSI, this paper proposes an adversarial selection fusion (ASF) method for the HSI–MSI fusion problem. Firstly, the unmixing based fusion (UF) method is adopted to dig out the spatial correlation in MSI. Then, to acquire the spectral correlation in HSI, a band reconstruction-based fusion (BRF) method is proposed, regarding H2SI as the product of the optimized band image dictionary and reconstruction coefficients. Finally, spectral spatial quality (SSQ) index is designed to guide the adversarial selection process of UF and BRF. Experimental results on four real-world images demonstrate that the proposed strategy achieves smaller errors and better reconstruction results than other comparison methods.
Highlights
In remote sensing, optical spectral image plays an important role
The unmixing based fusion (UF) method can obtain the spatial correlation in multispectral image (MSI), but the spectral correlation contained in hyperspectral image (HSI) is ignored
The image consists of 540 × 420 pixels with a spatial resolution of 2.5 m/pixel, mainly including agricultural and urban areas
Summary
Optical spectral image plays an important role. Due to the limits of the imaging platform, acquisition devices are usually designed with a tradeoff between spatial and spectral information [1]. The sparse regularization widely used in spectral unmixing research was introduced into the HSI–MSI fusion problem [21,22], assuming that the number of endmembers at each pixel is small compared to the number in the endmember matrix. Another abundance regularization called vector-total-variation was proposed by Simões et al [23], which controlled the spatial distribution of subspace coefficients, where the subspace can be defined either by singular value decomposition (SVD) or by endmember spectral signatures. The conventional unmixing based fusion (UF) method is adopted to obtain a preliminary H2SI, which focuses on keeping the spatial correlation of pixels in MSI. We treated the original hyperspectral data as the reference fusion result, and firstly generated the MSI by Equation (2) and the low spatial resolution HSI by Equation (1)
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
