Dual-Branch Remote Sensing Infrared Image Super-Resolution

2026-04-11

AI Summary

This work addresses the challenge of remote sensing infrared image super-resolution, which demands simultaneous enhancement of clarity while preserving object contours, scene layout, and radiometric stability—a task complicated by the inherently weak texture and sensitivity to local sharpening in thermal imagery. To this end, the study proposes a novel dual-branch collaborative architecture that explicitly integrates a Transformer (HAT-L) with a state space model (MambaIRv2-L): the former excels at restoring fine local details, while the latter ensures global structural consistency. Complementary gains are further achieved through test-time local adaptation, octant-based self-ensemble, and isotropic fusion in image space. Evaluated on the NTIRE 2026 Challenge and Caltech synthetic datasets, the proposed method significantly outperforms single-branch counterparts in PSNR, SSIM, and overall composite scores.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multi-instance/Multi-view Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Remote sensing infrared image super-resolution aims to recover sharper thermal observations from low-resolution inputs while preserving target contours, scene layout, and radiometric stability. Unlike visible-image super-resolution, thermal imagery is weakly textured and more sensitive to unstable local sharpening, which makes complementary local and global modeling especially important. This paper presents our solution to the NTIRE 2026 Infrared Image Super-Resolution Challenge, a dual-branch system that combines a HAT-L branch and a MambaIRv2-L branch. The inference pipeline applies test-time local conversion on HAT, eight-way self-ensemble on MambaIRv2, and fixed equal-weight image-space fusion. We report both the official challenge score and a reproducible evaluation on 12 synthetic times-four thermal samples derived from Caltech Aerial RGB-Thermal, on which the fused output outperforms either single branch in PSNR, SSIM, and the overall Score. The results suggest that infrared super-resolution benefits from explicit complementarity between locally strong transformer restoration and globally stable state-space modeling.

Problem

• infrared image super-resolution
• remote sensing
• thermal imagery
• radiometric stability
• weak texture

Innovation

• dual-branch
• infrared super-resolution
• transformer
• state-space model
• image fusion