Abstract
The CIE standard colorimetric observer defines the color matching functions of the average human observer which is used in many industrial and academic fields for the computation of various colorimetric, perceptual, and appearance attributes. It is well-known that there is significant variation in the spectral response amongst color-normal observers. This can be problematic, for instance, in the case of modern narrow-band primary, wide-gamut, high-dynamic-range displays wherein variations in individual-observer color matching functions and display primary spectra can cause significant color perception differences amongst observers. For many color-sensitive applications, there is interest in determining individual-observer color matching functions but with minimal knowledge of field-of-view, age, state-of-adaptation, and other viewing conditions in the actual use-setting. However, direct measurement of individual color matching functions in actual use-settings is tedious and impractical for many reasons. Similarly, in the camera industry, estimation of individual device spectral sensitivities is often used for production-line per-module color calibration. One method extracts the most significant eigenvectors from relevant spectral sensitivity data. Then several test stimuli whose spectral characteristics are chosen to co-optimize orthogonality, dimensionality, and producibility considerations are presented to the observer (camera) and the observer response to each test stimulus is recorded. From the eigenvectors, test stimuli spectra, and individual-observer responses, the individual-observer spectral sensitivities can be computed. Since human color matching functions vary significantly only along a few dimensions, the proposed method is extended to efficiently determining individual-human-observer color matching functions with a few simple measurements and little knowledge about the user-setting or viewing conditions.