Yi Han , Ming Yin, Puhong Duan, and Pedram Ghamisi
Haze in remote sensing images often leads to faded color and contrast reduction, which impairs image analysis and limits further applications (e.g., in land cover mapping, object detection, and environmental monitoring). Here, we propose an edge-preserving filtering-based method for haze removal in remote sensing images. In the proposed method, the original multispectral image is decomposed into base layers (containing haze) and detail layers (containing spatial detail) by multi-scale guided filtering. Then, the optimized atmospheric scattering model is used to estimate the atmospheric light and transmission, and to remove haze in the base layers. Next, adaptive non-linear mapping is used to enhance spatial details of ground objects in the detail layers. Finally, the resulting image is reconstructed by fusing the dehazed base layers and the refined detail layers. Experiments on simulated and real images demonstrate that the proposed method provides state-of-the-art performance.
IEEE Geoscience and Remote Sensing Letters, in press. 2021-08-17.