I am interested in using colour maps to measure background matching. However, my understanding is that because my background ROIs contain far more colours than my smaller subject ROIs, I will always get a low value of overlap even if the subject colour map is completely contained within the background colour map. Is it possible to just measure how much of the subject overlaps with the background, ignoring the unique colors in the background?
Thank you very much!
Hi,
Indeed, colour maps, just as most (if not all) colour pattern metrics, are best considered as relative measures. Some would argue that even the delta-S threshold in the RNL model is best understood as a relative measure (or any perceptual colour space metric such as CIElab, etc.). I.e. it should best be used to compare parameter X in context A relative to parameter X in context B. So yes, indeed, that mechanistic understanding explains why you have low overlap. But that doesn’t stop you from comparing the overlap in one subject vs. the overlap in another.
On the second point: You could cluster your image using RNL clustering and see how much of your subject is composed of colour pattern elements that it shares with the background.
Cheers,
Cedric