Speckle reduction is a preprocessing for synthetic aperture radar (SAR) image interpretation and application. With the advances of convolutional neural network (CNN) models, excellent speckle filters have been continuously developed. However, supervised learning-based models suffer from a generalization deficiency due to the lack of clean SAR images. Additionally, other self-supervised learning (SSL)-based methods rely on multiple independent SAR images of the same scene to generate the filtered SAR image. In practice, these additional auxiliary datasets are not always available, which limits the application of these methods. To fulfill this gap, a novel self-supervised framework named Sublook2Sublook is proposed for single SAR images speckle filtering. The main contribution of this work lies in that a new theorem is founded which guarantee the cost function defined on the paired sublook SAR images is statistically equivalent to the supervised counterpart based on the speckled-clean SAR image pairs. Thereby, the sublook SAR images can be alternatively used for model training instead of the clean SAR images or additional auxiliary datasets. From this fashion, a complete self-supervised speckle filter is developed. First, sublook decomposition is performed in both azimuth and range directions. Then, optimal paired sublook images are selected based on a criteria of minimum L1 norm distance. Finally, the established self-supervised speckle filter can be trained with the paired sublook images. Extensive experimental studies are conducted with various SAR datasets in terms of different frequency bands and spatial resolutions from the Radarsat-2, COSMO-SkyMed, and ALOS-2 SAR satellites. Comparisons studies with four state-of-the-art despeckling methods confirm the superiority of the proposed method. The results demonstrate that the proposed Sublook2Sublook framework can better smooth speckles in homogeneous areas while well preserve image details.

Polarimetric inverse synthetic aperture radar (ISAR) can operate all-day and all-weather, making it crucial for space surveillance. The compact polarimetric mode balances hardware complexity and polarimetric information, which is commonly equipped with ISAR systems. Given the constraints of limited physical conditions, exploring ISAR image super-resolution is worthwhile. Currently, deep learning models have been employed for enhancing ISAR image super-resolution. However, the super-resolution performance is limited by local interpolation and the occurrence of artifacts. To address these limitations, this work presents a Non-Local Scene Awareness Network (NLSAN), which incorporates a non-local interpolation approach to capture global textures. Furthermore, a scene awareness scheme is established by integrating semantic and super-resolution information, concerning the varying levels of artifacts in different regions. The training process can be regulated by a designed penalty function to mitigate potentially generated artifacts. A dataset of compact polarimetric ISAR images of satellite targets is constructed for comparison analysis. The proposed NLSAN method yields more elaborate super-resolution results with fewer artifacts. Quantitative evaluations are also carried out using global and local indexes such as the Peak-Signal-to-Noise (PSNR), the image entropy, and the 3dB width of strong scatters. Compared with the typical state-of-the-art methods, the proposed approach achieves superior super-resolution performance, with an overall performance improvement of at least 9.2% and enhanced generalization capabilities.

Compared with the optical imaging system, polarimetric inverse synthetic aperture radar (ISAR) can work all-day and all-weather, which plays an important role in space surveillance. However, high-resolution (HR) ISAR images usually require large bandwidth and coherent integration angle, which is limited by the equipment’s physical conditions. In this vein, the super-resolution (SR) of ISAR images is of vital importance. At present, supervised learning methods are often used in image SR of computer vision. By constructing low-resolution (LR) and HR data pairs, the neural network can learn the mapping relationship between them. However, the low-frequency information in LR image data is less considered. In addition, to obtain different scales of SR reconstruction results, multiple network training repetitions are usually needed, which consumes time and hardware resources. Based on the idea of implicit neural representation, this paper constructs an implicit neural network representation framework for polarimetric ISAR image SR, which can obtain multiscale SR results through one training. A semi-supervised module is also constructed to make the network have the ability of supervised and unsupervised learning, which is conducive to mine and make better use of LR images. A polarimetric ISAR image SR dataset is constructed for satellite targets while four indexes are adopted for quantitative evaluation in global and local aspects. Experiments demonstrate that the proposed approach achieves better SR performance, where the PSNR index can be increased at least by 0.93dB.

Polarimetric synthetic aperture radar (PolSAR) is widely used in remote sensing applications due to its ability to obtain full-polarization information. Compared to the quad-pol SAR, the dual-pol SAR mode has a wider observation swath and is more common in most SAR systems. The goal of reconstructing quad-pol SAR data from the dual-pol SAR mode is to learn the contextual information of dual-pol SAR images and the relationships among polarimetric channels. This work is dedicated to addressing this issue, and a multiscale feature aggregation network has been established to achieve the reconstruction task. Firstly, multiscale spatial and polarimetric features are extracted from the dual-pol SAR images using the pretrained VGG16 network. Then, a group-attention module (GAM) is designed to progressively fuse the multiscale features extracted by different layers. The fused feature maps are interpolated and aggregated with dual-pol SAR images to form a compact feature representation, which integrates the high- and low-level information of the network. Finally, a three-layer convolutional neural network (CNN) with a 1 × 1 convolutional kernel is employed to establish the mapping relationship between the feature representation and polarimetric covariance matrices. To evaluate the quad-pol SAR data reconstruction performance, both polarimetric target decomposition and terrain classification are adopted. Experimental studies are conducted on the ALOS/PALSAR and UAVSAR datasets. The qualitative and quantitative experimental results demonstrate the superiority of the proposed method. The reconstructed quad-pol SAR data can better sense buildings’ double-bounce scattering changes before and after a disaster. Furthermore, the reconstructed quad-pol SAR data of the proposed method achieve a 97.08% classification accuracy, which is 1.25% higher than that of dual-pol SAR data.

Urban damage investigation is an important application for polarimetric synthetic aperture radar (SAR), which is capable of sensing the target scattering mechanism changes before and after a natural disaster. Quad-pol SAR with fully polarimetric acquisition capability can better sense the scattering mechanism changes. Meanwhile, dual-pol SAR with wider swath is suitable for large area monitoring. In this vein, this work dedicates to generating pseudo quad-pol SAR data from dual-pol SAR mode to partially reconstruct fully polarimetric information. The main contributions contain two aspects. Firstly, a multi-scale feature aggregation convolutional neural network (CNN) has been proposed to reconstruct quad-pol SAR data, which includes a feature extraction (FE) module to collect multi-scale features from dual-pol SAR data in spatial and polarimetric domain, and a feature translation (FT) network aggregated with attention modules to deeply fuse the stacked multi-scale features and map them to quad-pol SAR covariance matrices. Then, a urban damage level estimation approach has been established with reconstructed quad-pol SAR data based on polarimetric coherence pattern interpretation tool. Experimental studies have been carried out in terms of both pseudo quad-pol SAR data reconstruction and urban damage level estimation. Comparison results demonstrate that the proposed method achieves better quad-pol SAR data reconstruction accuracy and urban damage level estimation accuracy. Moreover, compared with the urban damage level estimated by real quad-pol SAR data, the proposed method can achieve 99.83% estimation consistency within 5% error tolerance.

Polarimetric synthetic aperture radar (PolSAR) is an important sensor for earth observation. Multi-temporal PolSAR images obtained by successive observations of the region of interest contain rich polarimetric–temporal–spatial information of the land covers, which has wide applications. Speckle filtering becomes a necessary pre-processing for many subsequent applications. Currently, it is common to filter multi-temporal PolSAR data by directly using a speckle filter developed for single SAR or PolSAR data. The cross-correlation between different time series contains rich information in multi-temporal PolSAR images. How to utilize complete polarimetric–temporal–spatial information becomes a large challenge to achieve more satisfied performances of speckle reduction and details preservation simultaneously. This work dedicates to this issue and develops a novel speckle filtering approach for multi-temporal PolSAR data by multi-dimensional information fusion. The core idea is to establish an adaptive and efficient strategy of similar pixel selection based on the similarity test of multi-temporal polarimetric covariance matrices. This similar pixel selection scheme fuses the complete information of multi-temporal PolSAR data. The sensitivity of the proposed scheme is demonstrated with several typical and challenging texture patterns. Then, an adaptive speckle filter is established specifically for multi-temporal PolSAR data. Intensive comparison studies are carried out with airborne UAVSAR datasets and spaceborne ALOS/PALSAR datasets. Quantitative investigations in terms of the equivalent number of looks (ENL) and the figure of merit (FOM) indexes demonstrate and validate the superiority of the proposed method.