Jul 2026· ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences· Vol XI-3-2026, pp. 797-803· 0 citations· 8 references
TL;DR
A comparative evaluation of multiple generative adversarial network architectures for translating SAR data into realistic red, green and blue (RGB) imagery in agricultural landscapes shows that paired image-to-image translation frameworks consistently outperform unpaired approaches.
Abstract
Abstract. Synthetic Aperture Radar (SAR) offers an all-weather alternative, and recent advances in deep generative models provide opportunities to reconstruct optical-like imagery directly from SAR data. In this study, we conduct a comparative evaluation of multiple generative adversarial network (GAN) architectures for translating SAR data into realistic red, green and blue (RGB) imagery in agricultural landscapes. The models were trained using ALOS-2/PALSAR-2 quad-polarimetric (quad-pol) data. A distinctive feature of our work is the evaluation of not only backscatter coefficients (Gamma nought) but also polarimetric parameters derived from quad-pol decompositions, including the generalised Freeman–Durden, H/A/Alpha, and Yamaguchi four-component methods. The comparative results show that paired image-to-image translation frameworks consistently outperform unpaired approaches. In particular, paired methods such as feature-guiding GAN and pix2pixHD, achieved high similarity to PlanetScope reference imagery, with mean structural similarity index values exceeding 0.98 across all SAR inputs. In contrast, unpaired approaches demonstrated more variable performance depending on the input features. Notably, PUT showed significant improvement when H/A/Alpha or Yamaguchi decompositions were used, whereas Freeman–Durden produced results comparable to Gamma nought. The performance gap between paired and unpaired frameworks was most evident in heterogeneous landscapes, such as areas with adjacent grasslands and forests. These findings demonstrate the effectiveness of GAN-based translation from polarimetric SAR to RGB imagery for agricultural monitoring. The integration of polarimetric information adds value to unpaired learning schemes, and the ability to generate optical-like imagery under challenging observation conditions has strong potential for practical use in crop monitoring and assessment.
Abstract. Despite the rapid development of satellite sensors, acquiring high-resolution true-color images remains challenging. The reliance on paired panchromatic and multispectral images, as required by traditional techniques such as pansharpening, presents a significant challenge in scenarios where such data are not available. In this paper, a GAN-based PAN-to-RGB model is proposed to establish a novel framework for high-resolution, high-fidelity true-color images generation from remote sensing data. Symmetric luminance-color decoders are leveraged to learn the mapping between luminance, color, and spatial multi-scale features, effectively overcoming the color desaturation, inaccuracies, and distortion common in existing algorithms. By combining CNNs for local feature modeling and transformers for global feature modeling, high-resolution, high-fidelity RGB images were generated in CIELAB space. We conducted experimental validation on Gaofen-7 satellite data, and the results demonstrated that the proposed algorithm effectively mitigates the issues of desaturated, inaccuracy, and distorted colors, achieving significant improvements in key metrics such as FID, CF, and ΔCF.
Xiaowei He, Yingzi Xiong, Xiao Ling et al.· The International Archives o...· 0 citations
Abstract. Electro-optical (EO) imagery is intuitive but highly dependent on weather and illumination, whereas synthetic aperture radar (SAR) imagery provides reliable all-weather observations yet offers limited spectral information. To complement these modalities, recent studies have applied conditional generative adversarial network (cGAN)-based image-to-image translation to SAR-to-EO translation. However, side-looking SAR introduces spatial distortions such as foreshortening and layover that cause relative misalignment with EO imagery, undermining pixel-wise supervision and yielding structural discrepancies between translated outputs and reference EO imagery. In this study, we propose Geometry-Conditioned Pix2Pix (GC-Pix2Pix), which explicitly incorporates on the projected local incidence angle (PLIA) information derived from SAR imagery to better preserve structure and alignment in translated EO imagery. The method is based on Pix2Pix and comprises a two-branch generator and a PatchGAN discriminator. The generator consists of a main network that processes SAR polarimetric channels (VV, VH) and a conditioning subnetwork that extracts PLIA features. The subnetwork uses multi-layer convolutional blocks to capture local PLIA patterns, and the extracted features are then fused with features from the main branch and emphasized via a spatial attention module. For training and evaluation, we assembled a dataset over South Korea that combines Sentinel-1A/1C GRD VV/VH with PLIA and Sentinel-2B Level-2A RGB imagery. We compared GC-Pix2Pix against representative baselines. Across multiple image quality assessment metrics and complementary qualitative analyses, the proposed approach consistently improved SAR-to-EO translation performance. This study establishes a foundational framework for enhancing SAR data utility by synthesizing high-quality EO imagery as an alternative for continuous Earth observation.
Jinmin Lee, Minkyung Chung, Aisha Javed et al.· The International Archives o...· 0 citations
Satellite imagery often suffers from limited spatial resolution and, in many cases, high acquisition costs. These factors restrict their use in applications such as urban monitoring, land management, and wildlife studies. This work proposes an AI-based super-resolution approach that leverages high resolution aerial imagery to train a Generative Adversarial Network. Specifically, the ESRGAN (Enhanced Super-Resolution Generative Adversarial Network) architecture is adapted and trained using aerial orthophotos, enabling the transfer of learned spatial representations to low-resolution satellite images. The trained model is evaluated on satellite image patches at 2 and 4 super-resolution scales. Performance is assessed using structural, perceptual, and chromatic metrics, including SSIMY, MS-SSIM, LPIPS and CIEDE2000. The results show clear improvements, with increased sharpness, enhanced edge definition, and consistent reconstruction of urban structures and terrain features. From a quantitative perspective, the 2 scale achieves the best overall metric values, while the 4 scale maintains stable and meaningful performance despite the higher reconstruction difficulty. These findings demonstrate the feasibility of transferring super-resolution capabilities from aerial images to satellite imagery, even in the presence of spectral and geometric differences between acquisition domains. Overall, this study provides a solid foundation for the development of low-cost, AI-driven satellite image super-resolution models and outlines future research directions focused on dataset expansion, domain adaptation strategies, and sensor-specific architectural improvements.
Magda Alexandra Trujillo-Jiménez, Francisco Iaconis, Debora Pollicelli et al.· IEEE Latin America Transacti...· 0 citations
Abstract. Accurate and timely land cover mapping in heterogeneous metropolitan environments remains a fundamental challenge in Earth observation, particularly under conditions where optical imagery is compromised by cloud cover or seasonal atmospheric interference. This study presents a systematic evaluation of four state-of-the-art machine learning algorithms Random Forest (RF), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and a shallow Artificial Neural Network (ANN) for pixel-based land cover classification over the Istanbul metropolitan region using single-date ALOS-2 PALSAR-2 L-band Synthetic Aperture Radar (SAR) imagery. The methodological framework integrates dual-polarimetric backscatter coefficients (HH and HV) with Grey-Level Co-occurrence Matrix (GLCM) texture features, Land Parcel Identification System (LPIS) boundaries for reference data delineation, Bayesian hyperparameter optimization, and LightGBM-guided Recursive Feature Elimination (RFE) to establish a reproducible and computationally efficient classification pipeline. Among all tested configurations, LightGBM achieved the highest overall accuracy (OA = 85.1%, κ = 0.81) with a 10-feature subset identified through RFE, while XGBoost demonstrated the strongest performance for urban class discrimination. Bayesian optimization yielded statistically meaningful improvements over default configurations for all gradient-boosting models. The optimal feature count was found to be ten, with HV-derived texture features particularly Entropy and Contrast identified as the most discriminative predictors. These results confirm that systematic feature engineering and algorithm tuning are as critical as classifier selection in SAR-based land cover mapping and lay the foundation for scalable operational workflows applicable to rapidly urbanizing regions.
Melih Altay, B. Tavus, Fatih Fehmi Şi̇mşek et al.· The International Archives o...· 0 citations
Abstract. This paper focuses on cross-modal image matching between Synthetic Aperture Radar (SAR) and optical imagery, a longstanding challenge due to fundamental differences in imaging geometry and radiometry. Beyond applicational needs in satellite data fusion and downstream mapping, this study is motivated by the rapid advances in the field of Computer Vision. Thus, this work evaluates classical and modern learning-based feature matching methods on the renowned SpaceNet9 dataset using a unified evaluation framework. The results show that classical methods such as SIFT fail to produce reliable correspondences, while learning-based approaches, particularly MINIMA, significantly improve performance without additional retraining. However, matching accuracy is strongly influenced by scene structure and SAR-specific geometric effects, therefore robust SAR-optical correspondence remains an open challenge.
Constantin Günzel, M. Schmitt· The International Archives o...· 0 citations
Optical vegetation and soil indices are widely used in Earth observation, although their estimation is strongly affected by cloud coverage and illumination variability. Synthetic-aperture radar (SAR) has therefore attracted increasing interest as an alternative source for spectral index prediction. Most existing studies focus on directly estimating a single index from SAR observations. In this work, we investigate a more flexible formulation in which Sentinel-2 multispectral bands are first reconstructed from Sentinel-1 SAR data and subsequently used to derive multiple spectral indices. Experiments are conducted on the SEN12TP dataset, exploiting near-synchronous paired Sentinel-1 and Sentinel-2 acquisitions together with auxiliary elevation and land-cover information. Three SAR-to-multispectral reconstruction strategies are compared, namely, Efficient-UNet, Pix2Pix, and a conditional flow matching model. The resulting indices are then evaluated against those obtained through dedicated index-specific reconstruction models. The results show that Efficient-UNet achieves the best overall multispectral reconstruction performance among the evaluated architectures. Moreover, indices derived from reconstructed multispectral bands achieve performance comparable to dedicated index-specific models while offering substantially greater flexibility, as multiple indices can be computed within a single framework without retraining task-specific models. At the same time, the experiments highlight important intrinsic limitations of SAR-based spectral reconstruction. Although the reconstructed products preserve the large-scale spatial organization of the scenes, they do not fully recover fine spectral and vegetation-sensitive details. Consequently, SAR-derived spectral indices should be regarded as approximate proxies of optical observations rather than direct substitutes, particularly in applications requiring accurate biophysical interpretation.
Mirko Paolo Barbato, Roberto Cilli, Paolo Napoletano et al.· Remote Sensing· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
