遥感影像超分辨率重建降质模型研究
Study on Quality Reduction Model for Super-Resolution Reconstruction of Remote Sensing Images

遥感影像通过从航空器或卫星获取地表的图片，为环境监测与管理、农业、城市规划与管理、灾害管理和应急响应、国防和安全、科学研究等多个领域提供了极其宝贵的信息。目前，大部分的遥感影像超分辨率降质模型都倾向于使用单一的双三次降采样技术来模拟图像损失，或者采用传统的一阶退化模型，但这种方式并不能较好地模拟遥感图像在真实环境中的退化情况。为了更准确地模拟遥感图像在实际环境中的退化行为，本研究构建了一个高阶退化模型。这一模型在提升遥感图像重建质量方面表现出色，能更精确地复原遥感图像，并增强了网络模型处理各种类型的遥感图像时的泛化性能。
Remote sensing images provide extremely valuable information for many fields, such as environmental monitoring and management, agriculture, urban planning and management, disaster management and emergency response, national defense and security, and scientific research, by acquiring pictures of the ground surface from aircraft or satellites. Currently, most of the super-resolution quality degradation models for remote sensing images tend to use a single double-three times downsampling technique to simulate the image loss, or adopt the traditional first-order degradation model, but this approach does not better simulate the degradation of remote sensing images in the real environment. In order to more accurately simulate the degradation behavior of remote sensing images in real environments, a higher-order degradation model is constructed in this study. This model performs well in improving the quality of remote sensing image reconstruction, recovers remote sensing images more accurately, and enhances the generalization performance of the network model when dealing with various types of remote sensing images.