Creating, Interpreting and Rating Harmonic Colour Palettes Using a Cognitively Inspired Model

Abstract

This paper presents a cognitively inspired qualitative theory, \(QCharm\), which defines five operators for colour combination based on the qualitative colour descriptor (QCD) and applies these operators to recommend palettes of harmonic colours. Machine learning techniques have been applied to learn the QCD colour coordinates in Kobayashi’s colour space, in order to assign the resulting \(QCharm\) harmonic-colour palettes to cognitive keywords representing a feeling or a lifestyle. Furthermore, a regression model has been implemented to learn users’ preferences based on the COLOURlovers dataset. The resulting model is used as an additional criterion for recommendation. The resulting cognitive system can recommend (i) colour palettes using keywords on feelings/lifestyle, and (ii) colour palettes using the learnt user’s preference model. As an example of the practical applicability of the model, a web application, the \(QCharm\) tool, has been implemented to provide recommendations to users in an interactive way. The \(QCharm\) tool can also extract colour palettes from digital images and assign a cognitive adjective to describe colour combinations, to serve as a starting point for the design process.