NTIRE 2025 Challenge on RAW Image Restoration and Super-Resolution

2025-06-02

AI Summary

This work addresses joint image restoration and super-resolution (SR) in the RAW domain: simultaneously performing deblurring, denoising, and 2× upsampling on Bayer-pattern RAW images under unknown noise and blur degradations. We propose the first end-to-end deep network specifically designed for RAW-domain processing, featuring a noise-blind deblurring module, a Bayer-aware upsampler, and a photorealistic loss function—overcoming inherent limitations of conventional RGB-domain modeling. As the inaugural NTIRE 2023 Challenge on RAW Joint Restoration and Super-Resolution, we establish a standardized benchmark and a realistic degradation dataset with physically grounded noise and motion blur. The challenge attracted 230 registered teams, of which 45 submitted valid entries. Multiple quantitative metrics achieved new state-of-the-art (SOTA) performance, setting a new benchmark for low-level RAW vision tasks and advancing deep learning-based pre-ISP processing.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

This paper reviews the NTIRE 2025 RAW Image Restoration and Super-Resolution Challenge, highlighting the proposed solutions and results. New methods for RAW Restoration and Super-Resolution could be essential in modern Image Signal Processing (ISP) pipelines, however, this problem is not as explored as in the RGB domain. The goal of this challenge is two fold, (i) restore RAW images with blur and noise degradations, (ii) upscale RAW Bayer images by 2x, considering unknown noise and blur. In the challenge, a total of 230 participants registered, and 45 submitted results during thee challenge period. This report presents the current state-of-the-art in RAW Restoration.

Problem

• Restore RAW images with blur and noise degradations
• Upscale RAW Bayer images by 2x with unknown noise and blur
• Advance state-of-the-art in RAW Image Restoration and Super-Resolution

Innovation

• Restores RAW images with blur and noise
• Upscales RAW Bayer images by 2x
• Addresses unknown noise and blur degradation