MFE-Net: a remote sensing image dehazing model for the visual system of a flight simulator
Wenyi Ge, Bo Wang, Qi Liu, Xiaolin Qin, Shihan Tan, Shengjie Wang, Xia Yuan · 2025
Obtaining high-quality remote sensing images is crucial in generating a three-dimensional terrain for flight simulators. However, due to the presence of haze and other impact factors, collected remote sensing images usually suffer from blurred details, which limits its applicability real-world applications. To address this issue a deep learning-based dehazing model is proposed to advance hazy remote sensing images to build flight simulator visual systems. The encoder-decoder architecture is proposed to perform the image dehazing, in which an improved feature extraction module is designed to extract both global and local semantic features across multi-resolutions. In addition, the pixel-pixel and region-region mechanism are innovatively designed to achieve deep mapping relationships and further enhance image dehazing. Experimental results demonstrate that the proposed model outperforms other comparative baselines, achieving high-confidence dehazing in multi-scene remote sensing images, and with good generality and applicability in multi-resolution flight simulators and multi-terrain remote sensing.