Overcoming Attention Drift: Homogeneity-Heterogeneity Guided Feature Aggregation for Low-Light Remote Sensing Image Enhancement

2026-08-06

AI Summary

Extreme low-light remote sensing images are highly susceptible to noise and illumination degradation. Existing methods often suffer from attention drift, leading to erroneous cross-boundary feature aggregation that causes structural blurring and color distortion. To address this, this work proposes the HALO framework, which formulates image enhancement as a guided feature aggregation problem. HALO introduces, for the first time, a joint mechanism comprising a semantic homogeneity-induced positive bias and a pseudo-3D topological heterogeneity-based negative penalty. This design is realized through a Homogeneity–Heterogeneity Collaborative Attention Module (H2CAM) that effectively suppresses inter-boundary confusion. Evaluated on eight synthetic and real-world remote sensing datasets, the proposed method significantly improves edge sharpness and color fidelity while better preserving discriminative features critical for downstream Earth observation tasks.

Abstract

Restoring high-fidelity remote sensing imagery from extreme low-light degradation is indispensable for reliable Earth observation and downstream machine vision. However, under severe noise and illumination corruption, existing methods suffer from attention drift, erroneously aggregating features across distinct physical boundaries and causing severe structural blurring and color distortion. To address this, we propose HALO, a dual-prior-driven enhancement framework that formulates enhancement as a guided feature aggregation problem driven by foundation model priors. Specifically, an illumination-invariant semantic prior provides regional homogeneity as a positive bias for content-consistent aggregation, while a pseudo-3D topological prior provides boundary heterogeneity as a negative penalty to strictly prevent cross-boundary confusion. To cooperatively incorporate these two priors, we propose a Homogeneity-Heterogeneity Cooperative Attention Module (H2CAM) to resolve feature conflicts during cross-modal prior fusion. Extensive experiments demonstrate that HALO achieves state-of-the-art performance across 8 challenging synthetic and real-world remote sensing benchmarks, significantly improving physical boundary sharpness and color fidelity while maximizing the preservation of discriminative features for downstream Earth observation tasks.

Problem

• attention drift
• low-light remote sensing
• feature aggregation
• structural blurring
• color distortion

Innovation

• attention drift
• feature aggregation
• homogeneity-heterogeneity prior
• low-light enhancement
• remote sensing