Modelling inter-colour regions of Colour Naming Space

Although colour vision has common characteristics across humans, there is no universal agreement on how the colour space should be segmented to represent the names assigned by observers from different cultural backgrounds. To this respect, anthropologists and linguists have identified a limited set of eleven universal colour categories (names), that are present in most evolved languages, which could form the basis for a cross-cultural segmentation. In this work we present a mathematical formulation of fuzzy-sets to model this colour name assignment task, with a special focus on the inter-colour regions. In our model, the CIELab colour space is divided into eleven basic categories using a Triple Sigmoid as the fuzzy sets basis. To adjust the model’s parameters, psychophysical stimuli was created (in the CIE Lab space) from pairs of isoluminant colours belonging to different neighbouring categories and the colours in between. These were presented on a calibrated CRT monitor (14-bit x 3 precision) and observers indicated whether the colours belong to a category or another in a Yes/No discrimination paradigm. Our results show that inter-colour boundary regions are much less defined than expected and colour samples other than those near the focal colours (colours most representative of a given name) are needed to define the position and shape of boundaries between categories.