The Fourth Challenge on Image Super-Resolution ($\times$4) at NTIRE 2026: Benchmark Results and Method Overview

2026-04-16 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision · Depth: Expert, quick

Summary

The NTIRE 2026 image super-resolution (SR) challenge, held in conjunction with the NTIRE 2026 Workshop at CVPR 2026, focused on reconstructing high-resolution images from low-resolution inputs using a $\times$4 bicubic downsampling factor. This fourth iteration of the challenge aimed to benchmark effective SR solutions and analyze field advancements. It featured two distinct tracks: a restoration track, prioritizing pixel-wise fidelity and ranked by PSNR, and a perceptual track, emphasizing visual realism and evaluated via a perceptual score. The challenge attracted 194 registered participants, with 31 teams submitting valid entries. This report details the challenge's design, datasets, evaluation protocols, main results, and the methodologies employed by the participating teams.

Key takeaway

For research scientists developing image super-resolution models, understanding the NTIRE 2026 challenge's dual-track evaluation is crucial. You should consider optimizing your models for both pixel-wise fidelity (PSNR) and perceptual realism, as these represent distinct and important aspects of SR performance. Aligning your development with these established benchmarks can validate your approach against current state-of-the-art methods and guide future research directions.

Key insights

The NTIRE 2026 SR challenge benchmarks $\times$4 image super-resolution across fidelity and perceptual quality.

Principles

• Image SR requires balancing fidelity and perceptual quality.
• Bicubic downsampling is a standard LR image generation method.

Method

The NTIRE 2026 challenge evaluates super-resolution methods using two tracks: a restoration track (PSNR-based) and a perceptual track (perceptual score-based) for $\times$4 upscaling.

In practice

• Use PSNR for pixel-wise fidelity assessment.
• Employ perceptual scores for visual realism evaluation.

Topics

• NTIRE 2026
• Image Super-Resolution
• Bicubic Downsampling
• Pixel-wise Fidelity
• Perceptual Quality

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

Related on AIssential

• Beyond Visual Fidelity: Benchmarking Super-Resolution Models for Large-Scale Remote Sensing Imagery via Downstream Task Integration — Traditional SR metrics often fail to predict real-world performance in downstream remote sensing tasks.
• The Third Challenge on Image Denoising at NTIRE 2026: Methods and Results — The NTIRE 2026 challenge benchmarks unconstrained neural architectures for high-noise image denoising, prioritizing PSNR performance.
• TUDSR: Twice Upsampling-Diffusion for Higher Super-Resolution — TUDSR uses a two-stage, chunk-based diffusion framework to achieve high-quality super-resolution at resolutions up to 2048^2.
• NTIRE 2026: A Multi-Track Study in Image Restoration — NTIRE workshops benchmark image restoration models against real-world degradations using rigorous evaluation protocols.
• TOC-SR: Task-Optimal Compact diffusion for Image Super Resolution — TOC-SR creates efficient one-step super-resolution models via compact diffusion backbone discovery and distillation.
• Advancing Digital Twin Generation Through a Novel Simulation Framework and Quantitative Benchmarking — A novel simulation framework enables quantitative benchmarking of 3D photogrammetry pipelines using synthetic ground-truth data.

Open in AIssential →

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