Abstract | ||
---|---|---|
Due to the effect of the model mismatch error between the point spread function (PSF) and actual blur kernel, the performance of remote sensing image super-resolution (SR) is usually poor. In this paper, we propose a novel remote sensing image super-resolution method based on Lorentz fitting, to improve the reconstruction performance in actual application. Note that the actual blur kernel of remote sensing image is non-stationary and usually has non-smooth phenomenon in kernel edge. It is also asymmetric that the degree of blur is not same in radial and tangential directions, and blur is in a sense that the amount of blur depends on pixel locations in a sensor. This paper presents a flexible parametric blur kernel model based on a linear combination of Lorentz basic two-dimensional (2-D) patterns. The proposed model can provide flexible shapes for blur kernel with a different symmetry and non-smooth edge, which can model complicated blur due to various degradation factors accurately. Combining with the proposed PSF model and the optimized adaptive step size strategy, we proposed a remote sensing image super-resolution method to accelerate the convergence of super-resolution outputs. Experiment results have shown that the proposed method outperforms the other recent developed PSF model based remote sensing image super-resolution methods. |
Year | DOI | Venue |
---|---|---|
2022 | 10.1007/s11036-021-01870-x | Mobile Networks and Applications |
Keywords | DocType | Volume |
Remote sensing, Image super-resolution, Lorentz fitting, Asymmetric blur model, Parametric blurs estimation | Journal | 27 |
Issue | ISSN | Citations |
4 | 1383-469X | 0 |
PageRank | References | Authors |
0.34 | 25 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Huang Guoxing | 1 | 0 | 0.34 |
Yipeng Liu | 2 | 117 | 26.05 |
Weidang Lu | 3 | 309 | 55.86 |
Yu Zhang | 4 | 188 | 31.14 |
Hong Peng | 5 | 14 | 10.33 |