Observer metamerism (OM), which likely causes potential issues in high dynamic range (HDR) displays due to those formidable peak luminance and color gamut required, is examined in this work using simulations. The simulations focus on investigating how OM in HDR displays would vary with chromaticity gamut and peak luminance level changes, proposing a new OM index, OMM. The effects of additional factors noteworthy on OM, such as reference white level, age, and spectral characteristics, are discussed. Finally, a simple metric capable of predicting observer metamerism between displays with less computational complexity, OMMN, is introduced. The simulation results showed that observer metamerism magnitudes between displays tend to depend on the similarity in spectral bandwidth between paired displays, in addition to the fact that narrow‐band primary displays generally cause larger metameric failures. Besides, the effect of changes in peak luminance on observer metamerism was found to be relatively small, increasing OMM by 7%–8% when peak luminance doubles. Notably, the proposed efficient metric OMMN outperforms metrics based on spectral similarity by also factoring in inter‐observer variability. In addition, the proposed metric's performance with 10 categorical observers, OMMN, 10, is nearly as good as that with thousands more computations. OMMN, 10 is recommended as a reliable and efficient metric to evaluate OM between HDR displays.

This paper introduces a psychophysical experiment to investigate the image quality trade‐off relationship between peak luminance and the chromaticity area of the color gamut of a display device and models to predict equivalent image quality based on the experimental results. Our experimental results confirmed the hypothesis that the peak luminance required to maintain equivalent image quality tends to decrease as the color gamut area expands. At the same time, the relationship between peak luminance and color gamut strongly depends on image content. Based on the experimental results, two models are suggested. These models show good performance in terms of predicting the required luminance levels for given color gamut areas. Notably, the model built using an observer‐chosen region of interest (ROI) in an image slightly outperforms the model generated without this information.