基于可变形卷积与小波变换的多阶段遥感图像去噪算法

Multi-stage remote sensing image denoising algorithm based on deformable convolution and wavelet transform

遥感影像对记录地物目标信息至关重要,但其采集过程易受大气等外界因素影响,导致内在特性受到破坏。文章提出了一种深度学习遥感图像去噪算法,构建了一个新型多阶段特征优化网络,通过学习图像与噪声之间插值变化,可使用残差学习重建干净图像。经实验证明,该算法在图像去噪方面具有一定的有效性。

Remote sensing images are crucial for recording information about land targets,but their acquisition process is susceptible to external factors such as the atmosphere,which can damage their inherent characteristics.The article proposes a deep learning remote sensing image denoising algorithm and constructs a novel multi-stage feature optimization network.By learning interpolation changes between images and noise,residual learning can be used to reconstruct clean images.Experimental results have shown that this algorithm has certain effectiveness in image denoising.