NTIRE 2026 The 3rd Restore Any Image Model (RAIM) Challenge: Multi-Exposure Image Fusion in Dynamic Scenes (Track 2)

2026-04-10

AI Summary

This work addresses ghosting and artifacts in multi-exposure image fusion under dynamic scenes, which are caused by motion misalignment, illumination changes, and hand-held camera shake. To this end, the study introduces the first high dynamic range (HDR) image fusion benchmark tailored for real-world dynamic scenarios, providing aligned multi-exposure training and testing sequences. Comprehensive evaluation is conducted using standard metrics—PSNR, SSIM, and LPIPS—alongside assessments of alignment accuracy, ghosting suppression, detail preservation, and computational efficiency, thereby advancing end-to-end deep learning approaches. The associated challenge attracted 114 participating teams with 987 submissions, and the top-performing methods demonstrated significant improvements in robustness, perceptual quality, and artifact suppression for dynamic HDR fusion.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Computational Photography, Image & Video Synthesis
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

This paper presents NTIRE 2026, the 3rd Restore Any Image Model (RAIM) challenge on multi-exposure image fusion in dynamic scenes. We introduce a benchmark that targets a practical yet difficult HDR imaging setting, where exposure bracketing must be fused under scene motion, illumination variation, and handheld camera jitter. The challenge data contains 100 training sequences with 7 exposure levels and 100 test sequences with 5 exposure levels, reflecting real-world scenarios that frequently cause misalignment and ghosting artefacts. We evaluate submissions with a leaderboard score derived from PSNR, SSIM, and LPIPS, while also considering perceptual quality, efficiency, and reproducibility during the final review. This track attracted 114 participating teams and received 987 submissions. The winning methods significantly improved the ability to remove artifacts from multi-exposure fusion and recover fine details. The dataset and the code of each team can be found at the repository: https://github.com/qulishen/RAIM-HDR.

Problem

• multi-exposure image fusion
• dynamic scenes
• HDR imaging
• ghosting artefacts
• image misalignment

Innovation

• multi-exposure image fusion
• dynamic scenes
• HDR imaging
• ghosting artifact removal
• benchmark dataset