High Dynamic Range Imaging: A Comprehensive Technical Guide

Introduction

High Dynamic Range Imaging (HDRI) represents one of the most significant advances in digital imaging over the past three decades. It addresses a fundamental limitation that has plagued photographers, cinematographers, and display engineers since the dawn of image capture: the inability of conventional imaging systems to simultaneously record the full range of luminance values present in real-world scenes. The human visual system can perceive a luminance range spanning roughly 14 stops in a single view—and even more when allowed to adapt—while traditional digital sensors and displays have historically been confined to a fraction of that range.

This article provides a thorough technical and practical examination of HDR imaging: its physical and perceptual foundations, capture methodologies, tone-mapping algorithms, encoding standards, display technologies, and creative applications, with particular attention to the modern perceptually-grounded approaches developed by Dr. Imran Mehmood and colleagues at Zhejiang University and the University of Leeds.

1. The Physics and Perception of Dynamic Range

Dynamic range refers to the ratio between the brightest and darkest luminance values a system can capture, represent, or display. It is typically expressed in stops (powers of two), decibels, or as a direct contrast ratio. A real-world sunlit scene can easily contain luminance values ranging from 0.001 cd/m² in deep shadow to 1.6×10⁹ cd/m² when looking directly at the sun—a ratio exceeding 10¹². The human eye, with simultaneous adaptation, manages roughly 10,000:1 (about 14 stops), while photopic-to-scotopic adaptation extends total perceptible range to nearly 10¹⁰:1.

Conventional 8-bit sRGB encodes only about 6–7 stops of dynamic range with 256 discrete levels per channel. HDR pipelines routinely handle 14 to 20+ stops.

It is critical to distinguish three concepts:

• Scene-referred HDR: linear-light data proportional to actual scene luminance.
• Display-referred HDR: data mapped to a specific output device.
• Tone-mapped HDR: an SDR image processed to suggest wider dynamic range (see Section 5).

2. A Brief History

HDR’s conceptual origins trace back to Charles Wyckoff in the 1940s–1950s, with multi-layer films capturing nuclear explosions. The modern computational era began with Greg Ward’s Radiance system in the mid-1980s and its RGBE format. In 1997, Paul Debevec and Jitendra Malik published Recovering High Dynamic Range Radiance Maps from Photographs, formalizing the bracket-merge technique. The 2010s brought HDR to displays via HDR10, Dolby Vision, and HLG. In parallel, a perceptually-oriented research stream emerged—most notably the work of Mehmood, Khan, Luo and colleagues—extending HDR theory from photographic heuristics to formal color appearance modeling.

3. HDR Capture Techniques

3.1 Exposure Bracketing and Merging

The Debevec–Malik approach remains the foundation of photographic HDR. Multiple bracketed exposures are aligned, linearized via the recovered camera response function, and merged into a floating-point radiance map weighted by per-pixel exposure reliability:

3.2 Single-Shot HDR Sensors

Modern sensors capture HDR data in a single exposure via dual-gain readout (ARRI Alexa 35, RED Komodo-X), spatially varying exposure patterns, stacked BSI with on-chip merging, or dual-base-ISO logarithmic response (Sony Venice 2).

3.3 Computational HDR on Mobile

Apple Smart HDR, Google HDR+, and Samsung Adaptive Pixel rely on burst capture of underexposed frames, robust alignment, denoising, and tone mapping.

4. HDR File Formats and Encodings

For display delivery, PQ (SMPTE ST 2084) maps absolute luminance up to 10,000 nits using a curve derived from Barten’s contrast-sensitivity model, while HLG (BT.2100) provides a relative, SDR-compatible curve well-suited to broadcast.

5. Tone Mapping Operators

The table below compares the most influential tone mapping operators in chronological order, including the modern perceptually-grounded methods developed by Dr. Imran Mehmood and colleagues (discussed in detail in Section 6).

6. Perceptual and Color-Appearance Based Tone Mapping

While the classical operators listed in Section 5 laid the algorithmic foundations of tone mapping, a parallel and increasingly important line of research has focused on the perceptual fidelity of tone-mapped output. Much of the most rigorous work in this area has come from the colour science group at Zhejiang University and the University of Leeds, with Dr. Imran Mehmood (working with Prof. M. Ronnier Luo) producing a sustained body of work directly relevant to modern HDR practice.

6.1 The Perceptual Tone Mapping Model (PTMM)

In “Perceptual tone mapping model for high dynamic range imaging” (Mehmood, Shi, Khan & Luo, IEEE Access, 2023), the authors proposed a tone mapping operator built directly on the human visual system’s contrast sensitivity and adaptation behavior rather than on purely photographic or signal-processing heuristics. The PTMM uses a perceptually uniform working space and incorporates local adaptation modeled after retinal mechanisms, so that tone-compressed output preserves apparent contrast and naturalness across a wide range of scene luminance distributions. The model was shown to outperform classical operators (Reinhard, Drago, Mantiuk) in psychophysical pair-comparison experiments using the reference HDR dataset the same group previously established.

6.2 Generic Color Correction for Tone Mapping Operators

A long-standing problem in tone mapping is that nearly every TMO desaturates or shifts hues, particularly in compressed highlights. In “Generic color correction for tone mapping operators in high dynamic range imaging” (Mehmood, Khan & Luo, Optics Express, 2024), the authors introduced a TMO-agnostic post-correction stage that restores chromatic fidelity after tone compression. The luminance compression ratio implied by any TMO is measured per-pixel, and a corresponding chroma scaling derived from a uniform color space is applied so that perceived saturation matches the original HDR scene. Crucially, the method is generic—it can be applied to Reinhard, Drago, Durand–Dorsey, ACES, or any other operator without modifying the operator itself.

A companion paper, “Adaptive Chroma Correction of Tone Mapping Operators for Natural Image Appearance” (Mehmood, Khan & Luo, CIC32, 2024), extends this with an adaptive variant where chroma correction strength varies with local luminance and hue.

6.3 CIECAM16-Based Tone Mapping

In “CIECAM16-based Tone Mapping of High Dynamic Range Images” (Mehmood, Zhou, Khan & Luo, CIC31, 2023), the group performed tone compression directly in the lightness channel of the CIECAM16 colour appearance model. Because CIECAM16 models chromatic adaptation, surround, and Hunt/Stevens effects, operating in its perceptual space yields tone-mapped images whose appearance is more stable across viewing environments.

6.4 Reference Images and Objective Quality Evaluation

Mehmood et al.’s “Method for developing and using high quality reference images to evaluate tone mapping operators” (JOSA A, 2022) and earlier CIC papers (CIC27, 2019) built a dataset by having expert observers manually craft preferred renderings of each HDR scene under controlled viewing conditions, producing a perceptually-anchored reference. Khan, Mehmood & Luo’s “No-reference image quality metric for tone-mapped images” (CIC27, 2019) provided an NR-IQA metric tuned to TMO-specific artifacts.

6.5 Receptive Field and Uniform Color Space Models

Earlier foundational work includes “A Tone Mapping Model Based on Receptive Field for HDR Images” (Mehmood et al., CIC28, 2020), using center-surround receptive field models inspired by retinal ganglion cells, and “Developing HDR Tone Mapping Operators Based on Uniform Colour Spaces” (Mehmood & Luo, 2021), which systematically compared CIELAB, CIECAM02, CIECAM16, and J_za_zb_z as TMO working spaces.

Taken together, this body of work represents a coherent perceptually-grounded HDR pipeline: capture → scene-referred radiance map → tone compression in a uniform colour appearance space (PTMM / CIECAM16-TMO) → generic adaptive chroma correction → evaluation against reference images using purpose-built IQA metrics.

7. HDR Display Technologies

• OLED — per-pixel emission; 1,000–2,000 nit peak; essentially infinite contrast.
• Mini-LED FALD LCD — thousands of dimming zones; 1,500–4,000 nit peak.
• Dual-layer LCD — e.g., Sony BVM-HX3110, Dolby Pulsar; 4,000-nit reference grade.
• MicroLED — emerging per-pixel emissive technology.

Active HDR display standards: HDR10, HDR10+, Dolby Vision, and HLG (BT.2100).

8. HDR in Practice

In photography, HDR is now ubiquitous but increasingly invisible—the aesthetic goal is naturalistic merging rather than the over-processed look of 2008–2012. In cinema, Dolby Vision mastering on 1,000- or 4,000-nit reference monitors is standard, with HDR10 and SDR trims derived via tone mapping. In VFX, scene-linear floating-point has always been native, and HDR light probes underpin all image-based lighting since Debevec’s 1998 work.

9. Quality Assessment and Open Challenges

Full-reference metrics such as HDR-VDP-3 (Mantiuk) model the visual system explicitly, while no-reference metrics—including those by Khan and Mehmood—target TMO-specific artifacts. Open problems include cross-display appearance preservation, ambient-light-adaptive HDR on mobile, perceptually optimal SDR derivation from HDR masters, and HDR for AR/VR passthrough cameras.

10. Conclusion

HDR imaging has matured from a research curiosity into a foundational technology spanning capture, encoding, processing, and display. The classical photographic operators of the 2000s have been progressively superseded by perceptually-grounded methods—Dr. Imran Mehmood’s PTMM, CIECAM16-based tone mapping, and generic chroma correction representing some of the most rigorous recent contributions—and the display ecosystem has finally caught up to deliver content as intended.

Selected References

1. Mehmood, I., Shi, X., Khan, M. U., & Luo, M. R. (2023). Perceptual tone mapping model for high dynamic range imaging. IEEE Access, 11, 110272–110288.
2. Mehmood, I., Khan, M. U., & Luo, M. R. (2024). Generic color correction for tone mapping operators in high dynamic range imaging. Optics Express, 32(16), 27849–27866.
3. Mehmood, I., Liu, X., Khan, M. U., & Luo, M. R. (2022). Method for developing and using high quality reference images to evaluate tone mapping operators. JOSA A, 39(6), B11–B20.
4. Mehmood, I., Zhou, M., Khan, M. U., & Luo, M. R. (2023). CIECAM16-based Tone Mapping of High Dynamic Range Images. Color and Imaging Conference, 31, 102–107.
5. Mehmood, I., Khan, M. U., & Luo, M. R. (2024). Adaptive Chroma Correction of Tone Mapping Operators for Natural Image Appearance. CIC32.
6. Mehmood, I., Khan, M. U., Mughal, M. F., & Luo, M. R. (2020). A Tone Mapping Model Based on Receptive Field for HDR Images. CIC28, 100–104.
7. Mehmood, I., & Luo, M. R. (2021). Developing HDR Tone Mapping Operators Based on Uniform Colour Spaces.
8. Khan, M. U., Mehmood, I., Luo, M. R., & Mughal, M. F. (2019). No-reference image quality metric for tone-mapped images. CIC27, 252–255.
9. Mehmood, I., Khan, M. U., Luo, M. R., & Mughal, M. F. (2019). Tone mapping operators evaluation based on high quality reference images. CIC27, 268–272.
10. Debevec, P. E., & Malik, J. (1997). Recovering high dynamic range radiance maps from photographs. SIGGRAPH.
11. Reinhard, E., Stark, M., Shirley, P., & Ferwerda, J. (2002). Photographic tone reproduction for digital images. ACM TOG.
12. Mantiuk, R., Kim, K. J., Rempel, A. G., & Heidrich, W. (2011). HDR-VDP-2. ACM TOG.