Deep Metric Learning for Color Differences

Numerous attempts have been made to define a color space and a color distance metric that would closely resemble the human color vision. The uniformity has been the main challenge, the human vision system is more sensitive to some colors while less sensitive to others. A distance given by an ideal metric would match the color difference seen by the human vision system. This study attempts to define such a metric utilizing the spectral data and the available information on the distinguishable colors. Deep neural networks are used in metric learning for modeling the color space and the metric. The resulting metric is then tested against the standard CIEDE2000 metric. DNNs are also used to project spectral data onto a new color space. The results indicate that the new color space with the Euclidean metric is more perceptually uniform than the standard LAB color space with the CIEDE2000 metric. The new metric enables better understanding about the human vision system and measuring the color differences.