-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Welcome to the DTUTMO (DTU Tone Mapping Operator) wiki! This comprehensive documentation covers everything you need to know about this biologically-inspired HDR tone mapping framework.
DTUTMO is a physiologically-accurate computational framework for converting high dynamic range (HDR) imagery to display-referred output. Unlike simple compression operators, DTUTMO simulates the complete human visual processing pipeline, from optical phenomena through retinal processing to perceptual appearance.
- 🧬 Biologically Grounded: Models optical, photoreceptor, neural, and cognitive processing stages
- 🎨 Advanced Color Appearance: Novel DTUCAM model with dual adaptation (σ + R_max)
- ⚡ Real-Time Performance: Explicit inverse formulas and hybrid display mapping
- 📊 Extreme Dynamic Range: Handles 0.001 to 100,000 cd/m² (9+ orders of magnitude)
- 🎯 Production Ready: No parameter tuning required, validated components
- 🔧 Flexible Architecture: Modular design with optional GPU acceleration
-
Hood Adaptation with Gain Control: Physiologically-accurate adaptation formula σ = k₁((O₁+I_a)/O₁)^m with adaptive response ceiling R_max = k₂[(O₂+p·I_a)/O₂]^(-1/2)
-
Proper Spectral Processing: Two-stage RGB→XYZ→LMS transformation with separate rod extraction accounting for Purkinje shift (507nm vs 555nm peaks)
-
Explicit Inverse Photoreceptor Model: O(1) closed-form inverse formulas eliminate iteration while maintaining full physiological accuracy
-
Hybrid Display Mapping: Gradient-adaptive blending uses fast approximation for 70-90% of pixels, accurate inverse only where needed
-
DTUCAM Color Appearance: First CAM with physiologically-consistent dual adaptation, explicit inverse, and 9 orders of magnitude range
- Installation & Setup - Install DTUTMO and dependencies
- Quick Start Guide - Process your first HDR image
- Basic Examples - Common usage patterns
- System Architecture - Pipeline overview and design principles
- Pipeline Stages - Detailed explanation of each processing stage
- Mathematical Foundations - Equations and theory
- Configuration Guide - Customize DTUTMO behavior
- Display Mapping Strategies - Whiteboard, full inverse, and hybrid methods
- Color Appearance Models - DTUCAM, XLR-CAM, and CIECAM16
- Performance Optimization - GPU acceleration and profiling
- Use Cases & Applications - HDR photography, VR/AR, medical imaging
- API Documentation - Complete class and function reference
- FAQ & Troubleshooting - Common questions and solutions
- Development Guide - Contributing to DTUTMO
- Python 3.10+
- NumPy 1.20+
- SciPy 1.7+
- 4GB RAM
- Python 3.11+
- PyTorch 2.5.1+ (GPU acceleration)
- CUDA-capable GPU with 4GB+ VRAM
- 8GB+ system RAM
- Linux (Ubuntu 20.04+, tested)
- macOS (10.15+)
- Windows 10/11 (via WSL or native)
Current Version: 2.1
Last Updated: October 2025
Python Support: 3.10, 3.11, 3.12
License: MIT
If you use DTUTMO in your research, please cite:
@software{dtutmo2025,
title = {DTUTMO: DTU Tone Mapping Operator},
author = {Soreze, Thierry Silvio Claude},
year = {2025},
institution = {Technical University of Denmark},
version = {2.1},
url = {https://github.com/your-org/dtutmo}
}- 📧 Email: tsor-at-dtu-dot-dk
- 💬 Discussions: GitHub Discussions
- 🐛 Bug Reports: Issue Tracker
This work builds upon decades of vision science research. We acknowledge the foundational contributions of:
- Hood & Finkelstein (1986) - Adaptation model
- Watson & Yellott (2012) - Pupil model
- CIE 180:2010 - Disability glare standard
- Vangorp et al. (2015) - Local adaptation
- Ashraf et al. (2024) - CastleCSF
Ready to get started? Head to the Getting Started guide!