Beyond Pixel Fidelity: Minimizing Perceptual Distortion and Color Bias in Night Photography Rendering

2026-04-30

AI Summary

This work addresses the challenges of perceptual distortion and color inaccuracies in night-time image rendering, which are often exacerbated by extreme luminance contrasts and the overemphasis on pixel-level fidelity in conventional approaches. To overcome these limitations, the authors propose pHVI-ISPNet, a novel framework that introduces the HVI color space into RAW-to-RGB conversion for the first time. The method integrates RAW-domain feature extraction, wavelet-based feature propagation, sample-adaptive dynamic loss weighting, and a feature distribution–guided color constancy constraint to jointly optimize detail preservation, exposure robustness, and color accuracy. Evaluated on the NTIRE 2025 Night Rendering dataset, pHVI-ISPNet achieves state-of-the-art performance in both CIEDE2000 color difference and LPIPS perceptual distortion metrics while maintaining strong pixel-wise fidelity.

Abstract

Night Photography Rendering (NPR) poses a significant challenge due to the extreme contrast between dark and illuminated areas in scenes, stemming from concurrent capture of severely dark regions alongside intense point light sources. Existing methods, which are mainly tailored for fidelity metrics, reveal considerable perceptual gaps and often detract from visual quality. We introduce pHVI-ISPNet, a novel RAW-to-RGB framework built on the robust HVI color space. Our network integrates four distinct key refinements: RAW-domain feature processing and Wavelet-based feature propagation to mitigate high-frequency detail loss; sample-based dynamic loss coefficients to ensure stable learning across varying exposure levels; and loss term based on feature distributions to maintain rigorous color constancy. Evaluations on the dataset introduced in the NTIRE 2025 challenge on NPR confirm our approach achieves competitive fidelity while establishing new state-of-the-art results in both CIE2000 color difference and LPIPS. This validates our perceptually-driven design for high-quality nighttime imaging.

Problem

• Night Photography Rendering
• Perceptual Distortion
• Color Bias
• Extreme Contrast
• Visual Quality

Innovation

• Perceptual Rendering
• RAW-to-RGB
• Color Constancy
• Wavelet-based Feature Propagation
• Dynamic Loss Coefficients