A methodology for perceptual image quality assessment of smartphone cameras – color quality

Abstract

This research project was undertaken to develop a procedure for evaluating perceptually-based pictorial image quality for smartphone camera captures. Tone quality, color quality, and sharpness and noise were evaluated in separate experiments. In each test, observers scaled overall quality and then the individual image quality characteristic of test images from a variety of smartphone cameras relative to an anchor image. Results were reported in 2016 on the individual image quality characteristics relative to overall quality as well as on the development of objective measurements that correlate with the visual ratings for sharpness and noise. (Farnand, et al., 2016) In this work, the visual ratings for color quality were assessed relative to objective measurements with the results from this analysis indicating that high correlations between the two can be achieved. The perceptual results correlated best with colorimetric information taken from the test images, rather than images of test charts that were captured under lab conditions, which did not necessarily contain colors representative of the colors important to the scenes. Results also indicated that contrast information was needed in addition to the colorimetric information in order to achieve high correlation between subjective and objective information for all scenes. Further, results for a beach scene suggest that sand may serve as a useful memory color for predicting device capture perceptual performance.