The author uses CMC I:c (and somewhat incorrectly claims it's also called Delta E, which normally refers to one of the CIELAB metrics). Then he notes the asymmetry of that quasimetric, and a workaround -- averaging the C1,C2 and C2,C1 distances.

I wonder if perhaps he would be better served by simply using the actual "Delta E" metric, specifically ΔE*₀₀, which doesn't have this symmetry problem?

PS. Sexy interactive color matrix table. Small error in the text below -- values >=40 are highlighted, not >45.