NTIRE 2026 Low-light Enhancement: Twilight Cowboy Challenge

2026-08-10

AI Summary

This work addresses the visual degradation in low-light multi-frame smartphone raw images caused by hand-induced geometric misalignment, high noise levels, and mixed illumination. To tackle these challenges, the authors propose a raw-domain multi-frame fusion method that jointly handles motion misalignment and complex illumination-dependent noise without requiring pre-alignment. A large-scale dataset encompassing diverse real-world indoor and outdoor low-light scenes is introduced, along with a three-stage evaluation protocol. Blind assessment is conducted via the CodaBench platform using a private test set. Experimental results demonstrate that ten participating teams surpassed the baseline, with the top-performing method achieving a 6.49 dB gain in PSNR and a 0.0101 improvement in SSIM, setting a new state-of-the-art for burst-mode low-light image enhancement.

Abstract

This paper presents a review of the NTIRE 2026 Low-light Enhancement: Twilight Cowboy Challenge. The objective of the competition was to merge a set of misaligned smartphone images in the raw domain, captured in low-light conditions, into a single, clean image. Introduced setup simultaneously addresses two problems of low-light photography: visual degradations such as high noise and mixed scene illuminants, and the geometric inconsistencies caused by hand movement during multi-frame capture. To advance research in low-light and nighttime computational photography, a challenging dataset was collected comprising 585 real-world scenes, spanning indoor low-light and outdoor nighttime conditions, for training and benchmarking participant solutions. The competition employed a three-stage evaluation protocol: automatic validation via the CodaBench platform in stages one and two, followed by blind assessment on a private test set for the final ranking. Ten teams surpassed the established baseline, achieving improvements of up to +6.49 dB in PSNR and +0.0101 in SSIM, thereby establishing new state-of-the-art performance for burst-based low-light image enhancement. These results demonstrate significant progress in handling real-world noise, motion, and illumination variability in the low-light setting. Comprehensive results, leaderboards, and additional information are publicly available at https://nightimaging.org.

Problem

• low-light enhancement
• image alignment
• burst imaging
• noise reduction
• computational photography

Innovation

• low-light enhancement
• burst image fusion
• raw domain alignment
• computational photography
• real-world dataset