Decouple and Rectify: Semantics-Preserving Structural Enhancement for Open-Vocabulary Remote Sensing Segmentation

arXiv cs.CV / 4/3/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper addresses open-vocabulary remote sensing segmentation by noting that CLIP’s global, language-aligned visual features often underperform on fine structural boundary delineation.
It proposes DR-Seg, a decouple-and-rectify framework that splits CLIP feature channels into semantics-dominated vs structure-dominated subspaces to apply DINO-based structural enhancement without harming language-aligned semantics.
A prior-driven graph rectification module injects high-fidelity structural priors under DINO guidance to produce a refined branch for better spatial delineation.
An uncertainty-guided adaptive fusion module combines the refined DINO/rectified branch with the original CLIP branch dynamically for final predictions.
Experiments on eight remote sensing benchmarks show that DR-Seg achieves state-of-the-art performance, reflecting improved boundary quality while preserving open-vocabulary semantic grounding.

Abstract

Open-vocabulary semantic segmentation in the remote sensing (RS) field requires both language-aligned recognition and fine-grained spatial delineation. Although CLIP offers robust semantic generalization, its global-aligned visual representations inherently struggle to capture structural details. Recent methods attempt to compensate for this by introducing RS-pretrained DINO features. However, these methods treat CLIP representations as a monolithic semantic space and cannot localize where structural enhancement is required, failing to effectively delineate boundaries while risking the disruption of CLIP's semantic integrity. To address this limitation, we propose DR-Seg, a novel decouple-and-rectify framework in this paper. Our method is motivated by the key observation that CLIP feature channels exhibit distinct functional heterogeneity rather than forming a uniform semantic space. Building on this insight, DR-Seg decouples CLIP features into semantics-dominated and structure-dominated subspaces, enabling targeted structural enhancement by DINO without distorting language-aligned semantics. Subsequently, a prior-driven graph rectification module injects high-fidelity structural priors under DINO guidance to form a refined branch, while an uncertainty-guided adaptive fusion module dynamically integrates this refined branch with the original CLIP branch for final prediction. Comprehensive experiments across eight benchmarks demonstrate that DR-Seg establishes a new state-of-the-art.