Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Perceptual Assessment and Optimization of HDR Image Rendering

Peibei Cao, Rafal K. Mantiuk, Kede Ma

TL;DR

A family of HDR quality metrics, in which the key step is employing a simple inverse display model to decompose an HDR image into a stack of LDR images with varying exposures, which facilitate the alignment and prioritization of specific luminance ranges for more accurate and detailed quality assessment.

Abstract

High dynamic range (HDR) rendering has the ability to faithfully reproduce the wide luminance ranges in natural scenes, but how to accurately assess the rendering quality is relatively underexplored. Existing quality models are mostly designed for low dynamic range (LDR) images, and do not align well with human perception of HDR image quality. To fill this gap, we propose a family of HDR quality metrics, in which the key step is employing a simple inverse display model to decompose an HDR image into a stack of LDR images with varying exposures. Subsequently, these decomposed images are assessed through well-established LDR quality metrics. Our HDR quality models present three distinct benefits. First, they directly inherit the recent advancements of LDR quality metrics. Second, they do not rely on human perceptual data of HDR image quality for re-calibration. Third, they facilitate the alignment and prioritization of specific luminance ranges for more accurate and detailed quality assessment. Experimental results show that our HDR quality metrics consistently outperform existing models in terms of quality assessment on four HDR image quality datasets and perceptual optimization of HDR novel view synthesis.

Perceptual Assessment and Optimization of HDR Image Rendering

TL;DR

A family of HDR quality metrics, in which the key step is employing a simple inverse display model to decompose an HDR image into a stack of LDR images with varying exposures, which facilitate the alignment and prioritization of specific luminance ranges for more accurate and detailed quality assessment.

Abstract

High dynamic range (HDR) rendering has the ability to faithfully reproduce the wide luminance ranges in natural scenes, but how to accurately assess the rendering quality is relatively underexplored. Existing quality models are mostly designed for low dynamic range (LDR) images, and do not align well with human perception of HDR image quality. To fill this gap, we propose a family of HDR quality metrics, in which the key step is employing a simple inverse display model to decompose an HDR image into a stack of LDR images with varying exposures. Subsequently, these decomposed images are assessed through well-established LDR quality metrics. Our HDR quality models present three distinct benefits. First, they directly inherit the recent advancements of LDR quality metrics. Second, they do not rely on human perceptual data of HDR image quality for re-calibration. Third, they facilitate the alignment and prioritization of specific luminance ranges for more accurate and detailed quality assessment. Experimental results show that our HDR quality metrics consistently outperform existing models in terms of quality assessment on four HDR image quality datasets and perceptual optimization of HDR novel view synthesis.
Paper Structure (15 sections, 11 equations, 8 figures, 4 tables)

This paper contains 15 sections, 11 equations, 8 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. HDR Quality Metrics
  4. Novel View Synthesis
  5. Proposed HDR Quality Metrics
  6. Inverse Display Model
  7. Quality Assessment Model
  8. Quality Assessment Validation
  9. Experimental Setups
  10. Results
  11. Perceptual Optimization Validation
  12. HDR Novel View Synthesis
  13. Experimental Setups
  14. Experimental Results
  15. Conclusion and Discussion

Figures (8)

  • Figure 1: Forward display model simulates the process of converting digital pixel values into physical light in luminances on display. An inverse display model provides the inverse mapping.
  • Figure 2: System diagram of the proposed family of HDR-IQA metrics. The default red arrow can be replaced by the optional green arrow, whose goal is to compensate for the possible luminance shifts between the reference and test HDR images, similar to the camera response function correction in eilertsen2021cheatHanji2022.
  • Figure 3: Decomposition of an HDR image with eight stops into three LDR images of different exposures using the inverse display model in Eq. \ref{['eq:IDM']}. $l_0$ and $l_1$ denote the minimum and maximum luminances in the log-scale. (a) indicates the positions of the sliding windows by Eq. \ref{['eq:position']}. (b)-(d) are the LDR images corresponding to Window $1$ to Window $3$, respectively.
  • Figure 4: LDR image stack generated by the inverse display model (in Eq. \ref{['eq:IDM']}) and the corresponding local weighting maps (by Eq. \ref{['eqn:weight']}) for the "Forest" scene.
  • Figure 5: Illustration of compensation for the luminance shifts through Eq. \ref{['eq:omeh']}. (a) LDR image stack generated from the reference HDR image. (b) LDR image stack generated from the HDR image by MaskHDR Santos2020with the same exposure values used in (a). (c) LDR image stack generated from the HDR image by MaskHDR Santos2020 with optimized exposure values.
  • ...and 3 more figures