Interpretation-Oriented Cloud Removal via Observation-Anchored Residual Flow with Geo-Contextual Alignment

2026-07-02 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Remote Sensing & Geospatial AI · Depth: Expert, quick

Summary

The Geo-Anchored Cloud Removal (GACR) framework is proposed to enhance optical remote sensing by ensuring faithful cloud removal (CR) reconstruction and robust interpretability for downstream tasks like semantic segmentation and change detection. Existing CR methods often neglect their impact on analytical tasks, causing semantic drift. GACR integrates Observation-Anchored Residual Flow (OAR-Flow), which redefines CR as a physically grounded residual inversion, anchoring the generative process to cloudy observations for fast, stable, and faithful results. Additionally, Geo-Contextual Prior Alignment (GCPA) is incorporated to constrain reconstruction within a semantic manifold, leveraging a Vision Foundation Model (VFM) to maintain spatial-semantic integrity of complex landscapes. Extensive experiments across six CR datasets and twelve downstream tasks demonstrate GACR's superior reconstruction quality and consistent improvement in downstream accuracy. The code is available on GitHub.

Key takeaway

For Computer Vision Engineers developing remote sensing applications, if your cloud removal pipeline impacts downstream semantic segmentation or change detection, you should evaluate GACR. This framework directly addresses semantic drift by ensuring both faithful reconstruction and robust interpretability, potentially improving the accuracy of your analytical tasks. Consider integrating GACR's OAR-Flow and GCPA components to maintain spatial-semantic integrity in complex landscapes.

Key insights

GACR ensures faithful cloud removal and robust interpretability by anchoring generation and aligning semantic context.

Principles

• Cloud removal must prioritize downstream interpretability.
• Anchoring generative processes to observations improves stability.
• Semantic manifold constraints preserve spatial-semantic integrity.

Method

GACR integrates Observation-Anchored Residual Flow (OAR-Flow) for physically grounded residual inversion and Geo-Contextual Prior Alignment (GCPA) using a Vision Foundation Model (VFM) to constrain reconstruction.

In practice

• Apply GACR for cloud removal in optical remote sensing.
• Utilize VFM-induced semantic manifolds to maintain landscape integrity.

Topics

• Cloud Removal
• Remote Sensing
• Semantic Segmentation
• Change Detection
• Vision Foundation Models
• Generative AI

Code references

• wzy6055/GACR

Best for: AI Scientist, Computer Vision Engineer, Research Scientist

Related on AIssential

• ATT-CR: Adaptive Triangular Transformer for Cloud Removal — Adaptive Triangular Transformer (ATT-CR) efficiently removes clouds from remote sensing images by combining linear triangular attention and feature-selective gating.
• GMBFormer: An NDVI-Guided Global Memory Bank Transformer for Urban Green-Space Extraction from Ultra-High-Resolution Imagery — GMBFormer improves urban green-space extraction by decoupling NDVI and using a global memory bank for semantic reuse.
• RSRCC: A Remote Sensing Regional Change Comprehension Benchmark Constructed via Retrieval-Augmented Best-of-N Ranking — RSRCC is a new benchmark for fine-grained, localized remote sensing change question-answering.
• Heterogeneous SAR-optical fusion for near-real-time land use and land cover mapping under cloud contamination: A novel framework and global benchmark dataset — CloudLULC-Net fuses SAR and cloud-contaminated optical data for robust LULC mapping, overcoming traditional cloud limitations.
• Semantic Smoothing via Novel View Synthesis for Robust SAR Image Classification — Semantic smoothing uses generative models to create semantically consistent views for robust classification, outperforming isotropic noise.
• RPC-GS: Gaussian Splatting with native RPC Rendering for Satellite Imagery — Native RPC integration in Gaussian Splatting significantly improves satellite imagery 3D reconstruction accuracy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.