arXiv subjects

Elise Colin

Publications and source records attributed to Elise Colin.

7 recordsLinked to original sources

arXiv2026

SARLO-80: Worldwide Slant SAR Language Optic Dataset 80cm

Multimodal foundation models have advanced rapidly thanks to large optical benchmarks, but comparable resources for synthetic aperture radar (SAR) remain limited. Existing SAR--optical datasets largely rely on low-resolution, intensity-only Ground Range Detected~(GRD) products and do not preserve complex-valued SAR measurements or native acquisition geometry, which restricts physically grounded multimodal learning. In particular, large-scale public datasets combining very-high-resolution (VHR) SAR SLC, aligned optical imagery, and natural-language descriptions are still lacking. We present a VHR SAR--optical--text dataset built from open-access Umbra spotlight acquisitions distributed as Sensor Independent Complex Data (SICD). From around 2,500 worldwide scenes (VV/HH, 20cm--2m native resolution), we standardize all SAR data to an 80cm slant-range grid via band-limited FFT resampling and tile the imagery into 1024 by 1024 patches. For each SAR patch, we retrieve a high-resolution optical tile and warp it into the SAR grid using local coordinate correspondences for local pixel-level alignment. We further generate three caption variants (SHORT/MID/LONG) per sample to support vision--language training and evaluation. Our dataset contains 119,566 triplets (complex and amplitude slant-range SAR patch, aligned optical patch, natural-language description) covering 257 locations across 72 countries and a broad range of land types and infrastructures. We release fixed train/validation/test splits and the full preprocessing and baseline code to enable reproducible benchmarks for multimodal alignment on cross-modal retrieval and conditional generation in native SAR geometry. The dataset is publicly available on the Hugging Face Hub at https://huggingface.co/datasets/ONERA/SARLO-80.

cs.CV↗

arXiv2025

Quantitative Comparison of Fine-Tuning Techniques for Pretrained Latent Diffusion Models in the Generation of Unseen SAR Images

We present a framework for adapting a large pretrained latent diffusion model to high-resolution Synthetic Aperture Radar (SAR) image generation. The approach enables controllable synthesis and the creation of rare or out-of-distribution scenes beyond the training set. Rather than training a task-specific small model from scratch, we adapt an open-source text-to-image foundation model to the SAR modality, using its semantic prior to align prompts with SAR imaging physics (side-looking geometry, slant-range projection, and coherent speckle with heavy-tailed statistics). Using a 100k-image SAR dataset, we compare full fine-tuning and parameter-efficient Low-Rank Adaptation (LoRA) across the UNet diffusion backbone, the Variational Autoencoder (VAE), and the text encoders. Evaluation combines (i) statistical distances to real SAR amplitude distributions, (ii) textural similarity via Gray-Level Co-occurrence Matrix (GLCM) descriptors, and (iii) semantic alignment using a SAR-specialized CLIP model. Our results show that a hybrid strategy-full UNet tuning with LoRA on the text encoders and a learned token embedding-best preserves SAR geometry and texture while maintaining prompt fidelity. The framework supports text-based control and multimodal conditioning (e.g., segmentation maps, TerraSAR-X, or optical guidance), opening new paths for large-scale SAR scene data augmentation and unseen scenario simulation in Earth observation.

cs.CV↗

arXiv2025

From Spaceborne to Airborne: SAR Image Synthesis Using Foundation Models for Multi-Scale Adaptation

The availability of Synthetic Aperture Radar (SAR) satellite imagery has increased considerably in recent years, with datasets commercially available. However, the acquisition of high-resolution SAR images in airborne configurations, remains costly and limited. Thus, the lack of open source, well-labeled, or easily exploitable SAR text-image datasets is a barrier to the use of existing foundation models in remote sensing applications. In this context, synthetic image generation is a promising solution to augment this scarce data, enabling a broader range of applications. Leveraging over 15 years of ONERA's extensive archival airborn data from acquisition campaigns, we created a comprehensive training dataset of 110 thousands SAR images to exploit a 3.5 billion parameters pre-trained latent diffusion model \cite{Baqu2019SethiR}. In this work, we present a novel approach utilizing spatial conditioning techniques within a foundation model to transform satellite SAR imagery into airborne SAR representations. Additionally, we demonstrate that our pipeline is effective for bridging the realism of simulated images generated by ONERA's physics-based simulator EMPRISE \cite{empriseem_ai_images}. Our method explores a key application of AI in advancing SAR imaging technology. To the best of our knowledge, we are the first to introduce this approach in the literature.

eess.IV↗

arXiv2025

Turning a Disposable Bronchoscope into a Dynamic Speckle Imaging Tool: Yes, It Works

Dynamic speckle imaging, typically used in laser-illuminated surface diagnostics, has proven valuable for assessing biological activity. In this work, we demonstrate its feasibility in an endoscopic context using a disposable bronchoscope. Despite technical limitations and aliasing artifacts, our preliminary results show discernible vascular structures, indicating potential for minimally invasive diagnostic applications. It is important to note that the imaging systems used in this study are designed primarily for clinical robustness and classical imaging, including single-use sterility, ease of handling, and real-time visualization, and not for scientific fidelity of visual data or computational post-processing. As such, they are not inherently suited to dynamic speckle analysis, which requires precise control over temporal acquisition parameters, linear response characteristics of the imaging sensor, and stable illumination conditions, particularly from the coherent laser source. Nevertheless, our results demonstrate that, even within these constraints, dynamic speckle imaging is indeed achievable. This opens the door to further adaptation and optimization of such clinical imaging tools for functional biomedical investigations.

physics.ins-det↗

arXiv2025

Distribution and Moments of a Normalized Dissimilarity Ratio for two Correlated Gamma Variables

We consider two random variables $X$ and $Y$ following correlated Gamma distributions, characterized by identical scale and shape parameters and a linear correlation coefficient $\rho$. Our focus is on the parameter: \[ D(X,Y) = \frac{|X - Y|}{X + Y}, \] which appears in applied contexts such as dynamic speckle imaging, where it is known as the \textit{Fujii index}. In this work, we derive a closed-form expression for the probability density function of $D(X,Y)$ as well as analytical formulas for its moments of order $k$. Our derivation starts by representing $X$ and $Y$ as two correlated exponential random variables, obtained from the squared magnitudes of circular complex Gaussian variables. By considering the sum of $k$ independent exponential variables, we then derive the joint density of $(X,Y)$ when $X$ and $Y$ are two correlated Gamma variables. Through appropriate varable transformations, we obtain the theoretical distribution of $D(X,Y)$ and evaluate its moments analytically. These theoretical findings are validated through numerical simulations, with particular attention to two specific cases: zero correlation and unit shape parameter.

math.ST↗

arXiv2024

Towards a Unified Formalism of Multivariate Coefficients of Variation -- Application to the Analysis of Polarimetric Speckle Time Series

This article primarily aims to unify the various formalisms of multivariate coefficients of variation, leveraging advanced concepts of generalized means, whether weighted or not, applied to the eigenvalues of covariance matrices. We highlight the existence of an infinite number of these coefficients and demonstrate that they are bounded. Moreover, we link the various coefficients of variation identified in the literature to specific instances within our unified formalism. We illustrate the utility of our method by applying it to a time series of polarimetric radar imagery. In this context, the coefficient of variation emerges as a key tool for detecting changes or identifying permanent scatterers, which are characterized by their remarkable temporal stability. The multidimensionality arises from the diversity of polarizations. The introduction of the various possible coefficients demonstrates how their selection impacts the detection of samples exhibiting specific temporal behaviors and underscores the contribution of polarimetry to dynamic speckle analysis.

physics.ins-det↗

arXiv2023

The speckle contrast extended to the polarimetric case: applications to radar and Laser images

This article proposes the application of various alternative definitions of the multivariate coefficient of variation parameter in two domains: radar polarimetric time series and dynamic polarimetric speckle. In the first case, the focus is on detecting permanent scatterers or changes, while in the second case, it involves calculating activity images. Our study demonstrates that most of these parameters offer added value in terms of signal-to-noise ratio improvement and enhancing contrast in specific regions. Furthermore, the concept of polarimetric multivariate coefficient of variation proves to be closely related to the degree of polarization.

eess.IV↗