Blind image deblurring is a long-standing challenge in remote sensing image restoration tasks. It aims to recover a latent sharp image from a blurry image while the blur kernel is unknown. To solve this problem, many image priors-based algorithms and learning-based algorithms have been proposed. However, most of these methods are based on a single blurry image. Due to the lack of high frequency information, the images restored by these algorithms still have some deficiencies in edge and texture details. In this work, we propose a novel deep learning model named Reference-Based Multi-Level Features Fusion Deblurring Network (Ref-MFFDN), which registers the reference image and the blurry image in the multi-level feature space and transfers the high-quality textures from registered reference features to assist image deblurring. Comparative experiments on the testing set prove that our Ref-MFFDN outperforms many state-of-the-art single image deblurring approaches in both quantitative evaluation and visual results, which indicates the effectiveness of using reference images in remote sensing image deblurring tasks. More ablation experiments demonstrates the robustness of Ref-MFFDN to the input image size, the effectiveness of multi-level features fusion network (MFFN) and the effect of different feature levels in multi-feature extractor (MFE) on algorithm performance.

Tomographic synthetic aperture radar (TomoSAR) is an advanced SAR interferometric technique to retrieve 3D spatial information. However, the standard deviation in the reconstructed elevation could be high due to the noise in the interferometric phases, which makes the denoising filter crucial before tomographic reconstruction. In this paper, we propose an unsupervised multichannel SAR interferometric phase denoising method based on the Convolution Neural Network (CNN). It utilizes the Weighted Least-Squares (WLS) regularization combining with the covariance of multichannel interferometric phases to minimize the standard deviation of phase noise, which leads to the accurate and complete TomoSAR reconstruction. This network is trained by real SAR images and the results of both simulated and real observations verify the effectiveness of our proposed method.

Blind deblurring of optical remote sensing images has been a longstanding challenge. In recent years, many learning-based deblurring algorithms have been greatly developed. However, these methods often suffer from losing image texture details and some artificial artifacts under low signalto- noise ratio (SNR) conditions. To tackle this challenge, we introduce an innovative end-to-end noise-robust blind deblurring algorithm based on the diffusion model joined with a denoising module and a wavelet-enhanced conditional embedding mechanism. Experiments have verified the effectiveness of our method. Compared to the image blind deblurring algorithms based on the diffusion model, the proposed algorithm demonstrates better performance in terms of quantitative metrics Peak Signal-to- Noise Ratio (PSNR) and Structural Similarity Index (SSIM). Compared to all the comparative algorithms, the proposed algorithm exhibits significant advantages in quantitative metrics Learned Perceptual Image Patch Similarity (LPIPS), Fr'echet Inception Distance (FID) and Natural Image Quality Evaluator (NIQE) and shows superior visual effects in restoring texture details in the restored images, especially in challenging low SNR conditions.

Sparse signal processing techniques, such as compressed sensing, are commonly employed in tomographic synthetic aperture radar (TomoSAR) imaging due to the sparsity present in the elevation direction. However, classical compressed sensing methods, such as ℓ ₁ norm regularization and orthogonal matching pursuit (OMP), suffer from the off-grid effect. Specifically, they discretize the elevation axis into multiple grids and assume that scatterers are located precisely on the grids, leading to reconstruction results that deviate from the true heights of the scatterers. Although gridless compressed sensing methods, such as atomic norm minimization (ANM), have achieved gridless reconstruction in specific scenarios, they face challenges such as the requirement of uniformly distributed baselines and large computational cost. In this paper, we propose a gridless compressed sensing method based on the alternate descent conditional gradient (ADCG) kernel for TomoSAR inversion, and compare it with ANM, iterative soft thresholding (IST), and OMP. We show through numerical simulations and experimental results that our proposed method is applicable not only to scenarios with uniformly distributed baselines, but also to scenarios with non-uniformly distributed baselines, and resolves the off-grid effect in both cases. Finally, we demonstrate the effectiveness of our proposed method by implementing gridless reconstruction using an actual dataset of urban buildings in Yuncheng.

Synthetic aperture radar (SAR) images are often affected by range ambiguity due to antenna sidelobe characteristics and pulse operating mechanism. The work of range ambiguity suppression focuses on both SAR system design and signal processing. On the one hand, ideas have been proposed to modify the transmitting and receiving waveforms to block the ambiguous energy, such as up and down chirp modulation, multiple elevation beams, and azimuth phase coding techniques. On the other hand, signal processing methods in echo and image domains have been used to reduce the range ambiguity energy, such as up and down chirp processing and sparse regularization methods. This article proposes a range ambiguity suppression method that combines sparse reconstruction and matched filter. There are four main steps: performing the sparse reconstruction to obtain the ambiguity area, estimating the ambiguity signal using the ambiguity area image and reconstruction matrix, separating the ambiguity signal from the whole signal to obtain the signal of the primary imaging area, and using the matched filter to the primary signal to obtain the main lobe area image. In this method, sparse reconstruction improves the accuracy of ambiguity signal estimation, and the matched filter ensures the efficiency of the imaging processing procedure. Simulation results show that the proposed method can effectively suppress range ambiguity. When the ambiguity energy is relatively strong, the focal underdetermined system solver algorithm can get better results than the classical orthogonal matching pursuit algorithm. Moreover, the weak targets and the details of the main lobe image are well preserved.