Interval estimation of color parameters from digital images

Abstract

Influence of the analog-to-digital conversion of an input signal from a sample onto the color rendition of a colorimetric device is studied. It is shown that irrespective of the light source choice, a specific systematic quantization error is introduced to the color rendition. We analyze the quantization error for a color photomatrix with a resolution of 8 bits per channel. An algorithm for finding the interval of values of the dominant wavelength and the color saturation of the sample is developed. It is shown that the similarity of two colors in a sample can be estimated by means of the Jaccard similarity coefficient. The spectral dependence of the intervals of color parameters, calculated from digital images, is found by means of a computing experiment. It is found that lower-brightness colors have a larger interval and are calculated with a lower accuracy. A hypothesis explaining the emergence of MacAdam ellipses on the chromaticity diagram is offered, namely the coding process of the visual neural signals can include procedures similar to the analog-to-digital conversion.