The amount of noise included in a hyperspectral image limits its application and has negative impact on classification, unmixing, target detection, so on. In images, because the intensity different bands is different, to better suppress high-noise-intensity preserve detailed information low-noise-intensity bands, denoising strength should be adaptively adjusted with bands. Meanwhile, same band, there exist spatial property regions, such as homogeneous regions edge or texture regions; reduce...

Pan-sharpening is a fundamental and significant task in the field of remote sensing imagery processing, which high-resolution spatial details from panchromatic images are employed to enhance resolution multispectral (MS) images. As transformation low MS image complex highly nonlinear, inspired by powerful representation for nonlinear relationships deep neural networks, we introduce multiscale feature extraction residual learning into basic convolutional network (CNN) architecture propose...

In the field of fusing multi-spectral and panchromatic images (Pan-sharpening), impressive effectiveness deep neural networks has been recently employed to overcome drawbacks traditional linear models boost accuracy. However, best our knowledge, existing research works are mainly based on simple flat with relatively shallow architecture, which severely limited their performances. this paper, concept residual learning introduced form a very convolutional network make full use high...

Because of the internal malfunction satellite sensors and poor atmospheric conditions such as thick cloud, acquired remote sensing data often suffer from missing information, i.e., usability is greatly reduced. In this paper, a novel method information reconstruction in images proposed. The unified spatial-temporal-spectral framework based on deep convolutional neural network (STS-CNN) employs combined with supplementary information. addition, to address fact that most methods can only deal...

Fusing satellite observations and station measurements to estimate ground-level PM2.5 is promising for monitoring pollution. A geo-intelligent approach, which incorporates geographical correlation into an intelligent deep learning architecture, developed PM2.5. Specifically, it considers distance spatiotemporally correlated in a belief network (denoted as Geoi-DBN). Geoi-DBN can capture the essential features associated with from latent factors. It was trained tested data China 2015. The...

The interactions between PM2.5 and meteorological factors play a crucial role in air pollution analysis. However, previous studies that have researched the relationships concentration conditions been mainly confined to certain city or district, correlation over whole of China remains unclear. Whether spatial seasonal variations exist deserves further research. In this study, were investigated 68 major cities for continuous period 22 months from February 2013 November 2014, at season, year,...

In this paper, to break the limit of traditional linear models for synthetic aperture radar (SAR) image despeckling, we propose a novel deep learning approach by non-linear end-to-end mapping between noisy and clean SAR images with dilated residual network (SAR-DRN). SAR-DRN is based on convolutions, which can both enlarge receptive field maintain filter size layer depth lightweight structure. addition, skip connections strategy are added despeckling model details reduce vanishing gradient...

Satellite aerosol products have been widely used to retrieve ground PM2.5 concentration because of their wide coverage and continuous spatial distribution. While more studies focused on the retrieval algorithms, foundation for retrieval—relationship between satellite optical depth (AOD) has not fully investigated. In this study, relationships AOD were investigated in 368 cities mainland China from February 2013 December 2017, at different temporal regional scales. Pearson correlation...

Hyperspectral image (HSI) denoising is a crucial preprocessing procedure to improve the performance of subsequent HSI interpretation and applications. In this paper, novel deep learning-based method for task proposed, by learning nonlinear end-to-end mapping between noisy clean HSIs with combined spatial-spectral convolutional neural network (HSID-CNN). Both spatial spectral information are simultaneously assigned proposed network. addition, multiscale feature extraction multilevel...

The existence of hybrid noise in hyperspectral images (HSIs) severely degrades the data quality, reduces interpretation accuracy HSIs, and restricts subsequent HSI applications. In this paper, spatial-spectral gradient network (SSGN) is presented for mixed removal HSIs. proposed method employs a learning strategy, consideration unique spatial structure directionality sparse spectral differences with additional complementary information effectively extracting intrinsic deep features Based on...

As a new earth observation tool, satellite video has been widely used in remote-sensing field for dynamic analysis. Video super-resolution (VSR) technique thus attracted increasing attention due to its improvement spatial resolution of video. However, the difficulty image alignment and low efficiency spatial–temporal information fusion make poor generalization conventional VSR methods applied videos. In this article, novel strategy temporal grouping projection an accurate module are proposed...

Recently, convolutional networks have achieved remarkable development in remote sensing image (RSI) super-resolution (SR) by minimizing the regression objectives, e.g., MSE loss. However, despite achieving impressive performance, these methods often suffer from poor visual quality with oversmooth issues. Generative adversarial (GANs) potential to infer intricate details, but they are easy collapse, resulting undesirable artifacts. To mitigate issues, this article, we first introduce...

Satellite video is an emerging type of earth observation tool, which has attracted increasing attention because its application in dynamic analysis. However, most studies only focus on improving the spatial resolution satellite imagery. In contrast, few works are committed to enhancing temporal resolution, and joint spatial-temporal improvement even less. The enhancement can not produce high-resolution imagery for subsequent applications, but also provide potentials clear motion dynamics...

Optical-flow-based and kernel-based approaches have been extensively explored for temporal compensation in satellite Video Super-Resolution (VSR). However, these techniques are less generalized large-scale or complex scenarios, especially videos. In this paper, we propose to exploit the well-defined difference efficient effective compensation. To fully utilize local global information within frames, systematically modeled short-term long-term discrepancies since observe that offer distinct...

Transformer-based method has demonstrated promising performance in image super-resolution tasks, due to its long-range and global aggregation capability. However, the existing Transformer brings two critical challenges for applying it large-area earth observation scenes: (1) redundant token representation most irrelevant tokens; (2) single-scale which ignores scale correlation modeling of similar ground targets. To this end, paper proposes adaptively eliminate interference irreverent tokens...

Mixed noise pollution in HSI severely disturbs subsequent interpretations and applications. In this technical review, we first give the analysis different noisy HSIs conclude crucial points for programming denoising algorithms. Then, a general restoration model is formulated optimization. Later, comprehensively review existing methods, from model-driven strategy (nonlocal mean, total variation, sparse representation, low-rank matrix approximation, tensor factorization), data-driven 2-D...

Existing assessments might have underappreciated ozone-related health impacts worldwide. Here our study assesses current global ozone pollution using the high-resolution (0.05°) estimation from a geo-ensemble learning model, with key focuses on population exposure and all-cause mortality burden. Our model demonstrates strong performance, achieving mean bias of less than -1.5 parts per billion against in-situ measurements. We estimate that 66.2% is exposed to excess for short term (> 30 days...