Hyper-and multi-spectral image fusion is an important technology to produce hyper-spectral and hyper-resolution images,which always depends on the spectral response function andthe point spread function.However,few wo...Hyper-and multi-spectral image fusion is an important technology to produce hyper-spectral and hyper-resolution images,which always depends on the spectral response function andthe point spread function.However,few works have been payed on the estimation of the two degra-dation functions.To learn the two functions from image pairs to be fused,we propose a Dirichletnetwork,where both functions are properly constrained.Specifically,the spatial response function isconstrained with positivity,while the Dirichlet distribution along with a total variation is imposedon the point spread function.To the best of our knowledge,the neural network and the Dirichlet regularization are exclusively investigated,for the first time,to estimate the degradation functions.Both image degradation and fusion experiments demonstrate the effectiveness and superiority of theproposed Dirichlet network.展开更多
For many years, the status of surface vegetation has been monitored by using polar-orbiting satellite imagers such as Moderate Resolution Imaging Spectroradiometer(MODIS). However, limited availability of clear-sky sa...For many years, the status of surface vegetation has been monitored by using polar-orbiting satellite imagers such as Moderate Resolution Imaging Spectroradiometer(MODIS). However, limited availability of clear-sky samples makes the derived vegetation index dependent on multiple days of observations. High-frequency observations from the geostationary Fengyun(FY) satellites can significantly reduce the influence of clouds on the synthesis of terrestrial normalized difference vegetation index(NDVI). In this study, we derived the land surface vegetation index based on observational data from the Advanced Geostationary Radiation Imager(AGRI) onboard the FY-4B geostationary satellite. First, the AGRI reflectance of visible band and near-infrared band is corrected to the land surface reflectance by the 6S radiative transfer model. The bidirectional reflectance distribution function(BRDF) model is then used to normalize the AGRI surface reflectance at different observation angles and solar geometries, and an angle-independent reflectance is derived. The AGRI surface reflectance is further corrected to the MODIS levels according to the AGRI spectral response function(SRF). Finally, the daily AGRI data are used to synthesize the surface vegetation index. It is shown that the spatial distribution of NDVI images retrieved by single-day AGRI is consistent with that of 16-day MODIS data. At the same time, the dynamic range of the revised NDVI is closer to that of MODIS.展开更多
基金the Postdoctoral ScienceFoundation of China(No.2023M730156)the NationalNatural Foundation of China(No.62301012).
文摘Hyper-and multi-spectral image fusion is an important technology to produce hyper-spectral and hyper-resolution images,which always depends on the spectral response function andthe point spread function.However,few works have been payed on the estimation of the two degra-dation functions.To learn the two functions from image pairs to be fused,we propose a Dirichletnetwork,where both functions are properly constrained.Specifically,the spatial response function isconstrained with positivity,while the Dirichlet distribution along with a total variation is imposedon the point spread function.To the best of our knowledge,the neural network and the Dirichlet regularization are exclusively investigated,for the first time,to estimate the degradation functions.Both image degradation and fusion experiments demonstrate the effectiveness and superiority of theproposed Dirichlet network.
基金Supported by the National Key Research and Development Program of China (2021YFB3900400)National Natural Science Foundation of China (U2142212 and U2242211)。
文摘For many years, the status of surface vegetation has been monitored by using polar-orbiting satellite imagers such as Moderate Resolution Imaging Spectroradiometer(MODIS). However, limited availability of clear-sky samples makes the derived vegetation index dependent on multiple days of observations. High-frequency observations from the geostationary Fengyun(FY) satellites can significantly reduce the influence of clouds on the synthesis of terrestrial normalized difference vegetation index(NDVI). In this study, we derived the land surface vegetation index based on observational data from the Advanced Geostationary Radiation Imager(AGRI) onboard the FY-4B geostationary satellite. First, the AGRI reflectance of visible band and near-infrared band is corrected to the land surface reflectance by the 6S radiative transfer model. The bidirectional reflectance distribution function(BRDF) model is then used to normalize the AGRI surface reflectance at different observation angles and solar geometries, and an angle-independent reflectance is derived. The AGRI surface reflectance is further corrected to the MODIS levels according to the AGRI spectral response function(SRF). Finally, the daily AGRI data are used to synthesize the surface vegetation index. It is shown that the spatial distribution of NDVI images retrieved by single-day AGRI is consistent with that of 16-day MODIS data. At the same time, the dynamic range of the revised NDVI is closer to that of MODIS.
