We recently proposed a convolutional neural network (CNN) for remote sensing image pansharpening obtaining a significant performance gain over the state of the art. In this paper, we explore a number of architectural and training variations to this baseline, achieving further performance gains with a lightweight network that trains very fast. Leveraging on this latter property, we propose a target-adaptive usage modality that ensures a very good performance also in the presence of a mismatch with respect to the training set and even across different sensors. The proposed method, published online as an off-the-shelf software tool, allows users to perform fast and high-quality CNN-based pansharpening of their own target images on general-purpose hardware.

In recent years, there has been a growing interest in deep learning-based pansharpening. Thus far, research has mainly focused on architectures. Nonetheless, model training is an equally important issue. A first problem is the absence of ground truths, unavoidable in pansharpening. This is often addressed by training networks in a reduced resolution domain and using the original data as ground truth, relying on an implicit scale invariance assumption. However, on full resolution images results are often disappointing, suggesting such invariance not to hold. A further problem is the scarcity of training data, which causes a limited generalization ability and a poor performance on off-training test images. In this paper, we propose a full-resolution training framework for deep learning-based pansharpening. The framework is fully general and can be used for any deep learning-based pansharpening model. Training takes place in the high-resolution domain, relying only on the original data, thus avoiding any loss of information. To ensure spectral and spatial fidelity, a suitable two-component loss is defined. The spectral component enforces consistency between the pansharpened output and the low-resolution multispectral input. The spatial component, computed at high-resolution, maximizes the local correlation between each pansharpened band and the panchromatic input. At testing time, the target-adaptive operating modality is adopted, achieving good generalization with a limited computational overhead. Experiments carried out on WorldView-3, WorldView-2, and GeoEye-1 images show that methods trained with the proposed framework guarantee a pretty good performance in terms of both full-resolution numerical indexes and visual quality.

Deep learning (DL) in remote sensing has nowadays become an effective operative tool: it is largely used in applications, such as change detection, image restoration, segmentation, detection, and classification. With reference to the synthetic aperture radar (SAR) domain, the application of DL techniques is not straightforward due to the nontrivial interpretation of SAR images, especially caused by the presence of speckle. Several DL solutions for SAR despeckling have been proposed in the last few years. Most of these solutions focus on the definition of different network architectures with similar cost functions, not involving SAR image properties. In this article, a convolutional neural network (CNN) with a multi-objective cost function taking care of spatial and statistical properties of the SAR image is proposed. This is achieved by the definition of a peculiar loss function obtained by the weighted combination of three different terms. Each of these terms is dedicated mainly to one of the following SAR image characteristics: spatial details, speckle statistical properties, and strong scatterers identification. Their combination allows balancing these effects. Moreover, a specifically designed architecture is proposed to effectively extract distinctive features within the considered framework. Experiments on simulated and real SAR images show the accuracy of the proposed method compared with the state-of-art despeckling algorithms, both from a quantitative and qualitative point of view. The importance of considering such SAR properties in the cost function is crucial for correct noise rejection and details preservation in different underlined scenarios, such as homogeneous, heterogeneous, and extremely heterogeneous.

We propose a new method for synthetic aperture radar (SAR) image despeckling, which leverages information drawn from coregistered optical imagery. Filtering is performed by patchwise nonlocal means, working exclusively on SAR data. However, the filtering weights are computed by taking into account also the optical guide, which is much cleaner than the SAR image, and hence more discriminative. To avoid injecting optical-domain information into the filtered image, an SAR-domain statistical test is preliminarily performed to reject right away any risky predictor. Experiments on two SAR-optical data sets prove the proposed method to suppress very effectively the speckle, preserving structural details, and without introducing significant filtering artifacts. Overall, the proposed method compares favorably with all the state-of-the-art despeckling filters, and also with our own previous optical-guided filter.

In the framework of deep learning for synthetic aperture radar (SAR) speckle reduction, the methods presented in the literature mainly focus on the definition of new architectures and cost functions for better catching and preserving the properties of a real SAR image. The achieved results are interesting and promising but with many left open issues. The main critical problem, shared by all the methods, is the construction of a training dataset. This is due to the lack of a noise-free reference. In this work, a comparison among different training approaches (synthetic, multitemporal, and hybrid) is carried out in order to analyze their benefits and drawbacks. Four convolutional neural network (CNN)-based methods have been trained with the three different datasets for their assessment. Results on real SAR images have been carried out showing the peculiarities of each training approach.

Interferometric Synthetic Aperture Radar is an effective and widely adopted tool for earth observation. Based on interferograms it is possible to infer several information about the observed area. Two main problems affecting the interferogram can limit its application: phase noise and phase wrapping. In this paper the attention is focused on the first issue. Several algorithms have been developed for interferogram restoration. Given the wide spread of Deep Learning (DL) in the framework of image processing, DL based algorithms have been proposed for interferogram denoising. Most of the efforts have been devoted in designing specific network architectures or training dataset, rather than on the definition of a specific cost function, well suited for the problem under investigation. The aim of this manuscript is to define a new multi-objective cost function, specifically thought for the interferograms restoration problem: the idea is to provide a cost function able to take into account multiple aspects of the data under investigation (i.e. multi-objective). The cost function is implemented within a Convolutional Neural Network and a specific realistic training dataset is built, to account the main characteristics of real interferograms. The final outcome of the paper is the proposal of a new robust and accurate interferometeric phase denoising algorithm (namely InSAR-MONet ), able to remove undesired noise and, at the same time, able to preserve important phase details. The assessment of the method is conducted on simulated and real datasets, comparing quantitatively and qualitatively InSAR-MONet with the state of the art interferometric denoising algorithms.