Xuran Hu

Preprint · 2026 New

Geometry-Calibrated Closed-Form Shrinkage for SAR Despeckling

Xuran Hu, Mingzhe Zhu, Djordje Stanković, Yujie Zhu, Zhenpeng Feng, Yifang Ban, Ljubiša Stanković

A deterministic, training-free SAR despeckling method with closed-form shrinkage and analytically calibrated settings across images and sensors.

Preprint · 2026 New

γ-Bridge: A Look-Parametric Diffusion Bridge

Xuran Hu, Yujie Zhu, Tengxi Wang, Jilong Li, Wufan Zhao

A physically interpretable diffusion bridge for controllable Gamma-noise restoration and zero-shot SAR despeckling.

Preprint · 2026 New

GeoHeight-Bench: Towards Height-Aware Multimodal Reasoning in Remote Sensing

Xuran Hu, Zhitong Xiong*, Zhongcheng Hong, Yifang Ban, Xiaoxiang Zhu, Wufan Zhao*

A benchmark for evaluating height-aware multimodal reasoning across elevation-informed remote sensing tasks.

IEEE Signal Processing Letters · 2025

Lightweight Attention-Enhanced Multi-scale Detector for Robust Small Object Detection in UAV

Haitao Yang, Yingzhuo Xiong, Dongliang Zhang, Xiai Yan*, Xuran Hu*

A lightweight attention-enhanced detector for robust small-object localization in UAV imagery.

Preprint

Causal Attribution of Above-ground Biomass Dynamics in Sweden via SHAP

Xuran Hu

A SHAP-based attribution study of the environmental drivers behind above-ground biomass dynamics across Sweden.

In preparation for Remote Sensing of Environment

Knowledge-Based Systems · 2025

Interpretable Multi-task SAR Image Processing via GAN-based Unsupervised Manipulation

Xuran Hu, Mingzhe Zhu*, Ziqiang Xu, Zhenpeng Feng, Ljubiša Stanković

An unsupervised GAN framework that learns interpretable manipulations for multiple SAR image-processing tasks.

TELFOR · 2024

Perturbation on Feature Coalition: Towards Interpretable Deep Neural Networks

Xuran Hu, Mingzhe Zhu*, Zhenpeng Feng, Miloš Daković, Ljubiša Stanković

A structured perturbation method that explains neural predictions through meaningful feature coalitions.

IEEE JSTARS · 2024

Re-perceive Global Vision of Transformer for RSI Weakly Supervised Object Localization

Xuran Hu, Mingzhe Zhu*, Zhenpeng Feng, Ljubiša Stanković

A Transformer-based approach that reuses global context to improve weakly supervised localization in remote sensing images.

Neural Networks · 2024

Manifold-based Shapley Explanations for High Dimensional Correlated Features

Xuran Hu, Mingzhe Zhu*, Zhenpeng Feng, Ljubiša Stanković

A manifold-aware Shapley method for explaining models with high-dimensional, strongly correlated features.

Neurocomputing · 2024

Unveiling SAR Target Recognition Networks: Adaptive Perturbation Interpretation

Mingzhe Zhu, Xuran Hu*, Zhenpeng Feng, Ljubiša Stanković

An adaptive perturbation framework for revealing how SAR target-recognition networks form their decisions.

IGARSS · 2024

Manifold-based Shapley for SAR Recognition Network Explanation

Xuran Hu, Mingzhe Zhu*, Yuanjing Liu, Zhenpeng Feng, Ljubiša Stanković

A manifold-based Shapley formulation for more faithful explanations of SAR recognition networks.

IGARSS · 2024

SAR Despeckling via Regional Denoising Diffusion Probabilistic Model

Xuran Hu, Ziqiang Xu, Zhihan Chen, Mingzhe Zhu*, Zhenpeng Feng, Ljubiša Stanković

A regional denoising diffusion model designed to restore SAR imagery corrupted by speckle noise.