Investigating and Reducing the Impairment of Point Spread Effect For Spatiotemporal Fusion Of Remote Sensing Imagery

Spatiotemporal fusion of remote sensing imagery is a technology aiming to provide the synthetic dense satellite image series with medium spatial resolution. Presently, many spatiotemporal fusion approaches have been proposed. However, for the purpose of easier modeling, most approaches ignore the point spread effect of images with different spatial resolutions in the fusion, which may lead to impairment of the fusion performance. To address this problem, we develop a new method to reduce the impairment for spatiotemporal fusion, called dePSE. Specifically, the dePSE utilizes an adaptive flat gaussian kernel to learn the point spread function between the medium and coarse resolution images from the base medium and coarse image pair, which is then used to decompose the coarse resolution surface changes and reconstruct the coarse resolution images with higher quality. Finally, the reconstructed ones will replace the original ones to achieve the fusion. To validate the necessity and effect of the dePSE method, our experiments firstly investigate the impairment of point spread effect for spatiotemporal fusion using the simulated dataset, then test its performance using the real Landsat-MODIS dataset. The experimental results indicate that the point spread effect will lead to serious impairment for spatiotemporal fusion, including spatial distortions and accuracy decrease, which should be taken into account when designing new spatiotemporal fusion approaches. On the other hand, the proposed dePSE method can successfully reinforce the fused images via reducing the impairment of the point spread effect.