HDR Imaging and Tone Mapping: From Photographic Operators to PQ and HLG

HDR imaging and tone mapping cover several related but distinct tasks: capturing or representing a wide scene-luminance range, mapping that information to a display, and adapting the result to different display capabilities. The history is not a simple transition from traditional operators to neural processing. Photographic tone-reproduction research, HDR image capture, video standards and modern display mapping developed along partly separate paths.

For the measurement of highlight and shadow range in camera systems, see dynamic range in digital cameras.

Early digital tone reproduction

In 2002, Erik Reinhard, Michael Stark, Peter Shirley and James Ferwerda published “Photographic Tone Reproduction for Digital Images” in ACM Transactions on Graphics. The work addressed mapping high-dynamic-range image data to the limited range of a print or display, drawing on photographic practice. It is an important research milestone, but not the beginning of all HDR imaging or tone mapping. See the paper and reference implementation.

HDR capture and display are different problems

HDR capture can involve exposure bracketing, sensor techniques or computational fusion to represent a wider range of scene luminance. Tone mapping then determines how that information is rendered for a target display. A display standard defines signal and display characteristics; it does not dictate one universal capture method. Keep the acquisition, image-referred processing and display-referred mapping stages distinct.

HDR television standardisation: PQ and HLG

ITU-R Recommendation BT.2100 specifies HDR television image parameters for production and international programme exchange using Perceptual Quantization (PQ) and Hybrid Log-Gamma (HLG). The first BT.2100 recommendation appeared in 2016, followed by revisions; the current in-force edition is BT.2100-3 (February 2025). PQ and HLG use different transfer functions and have different system assumptions. See the ITU-R BT.2100 recommendation.

For the camera-side colour profiles and transformations that feed managed imaging pipelines, see colour science in digital cameras.

ITU-R reports on operational HDR production discuss monitoring, signal levels and coexistence with SDR workflows. These are important because tone mapping is not just a per-image artistic choice: video production requires consistency over time, between displays and through the distribution chain. See ITU guidance on HDR television and the current BT.2408 operational-practices report.

Local and global tone mapping

Global operators apply a mapping based on the image or signal as a whole. Local methods adapt the mapping to spatial neighbourhoods and can preserve local contrast, but may introduce halos, edge artefacts or inconsistent results across adjacent frames. For video, temporal stability is essential: a mapping that changes abruptly from frame to frame can produce visible flicker even when individual frames look acceptable.

Where neural methods fit

Machine-learning methods can estimate tone curves, perform local adaptation, reconstruct missing highlight or shadow detail, or combine information from multiple frames. Their performance depends on training data, target display, scene distribution and evaluation method. They do not replace the need for defined signal standards or validated colour management. Avoid claiming that neural processing universally solves motion, clipping or display adaptation.

For the relationship between HDR capture, fusion and computational processing, see HDR imaging technology.

How to evaluate a tone-mapping system

  1. Define whether the input is scene-referred HDR, a camera-rendered image or an HDR video signal.
  2. Specify the target display, peak luminance, black level, colour volume and viewing environment.
  3. Use test material with highlights, shadows, saturated colours, gradients and moving scenes.
  4. Measure clipping, luminance mapping, colour error, local contrast and temporal stability.
  5. Document the algorithm, version, parameters and display transform.
  6. Compare against a declared reference and report both visual and objective limitations.

Conclusion

HDR history includes photographic tone-reproduction research, multi-exposure capture, display standards and later learned processing. These strands should not be collapsed into a single timeline of “neural HDR”. A technically accurate workflow identifies the stage being processed, uses the correct signal standard and validates the result on the intended display.