A predictive model of colour differentiation

• Main Author: Flatla, David Raymond
• Other Authors: McCalla, Gordon I.
• Format: Others
• Language: en
• Published: University of Saskatchewan 2008
• Subjects: atypical colour perception; assistive devices; human-computer interaction; colour adaptation tools; colour blindness; colour perception modeling; colour perception
• Online Access: http://library.usask.ca/theses/available/etd-12222008-114824/

The ability to differentiate between colours varies from individual to individual. This variation is attributed to factors such as the presence of colour blindness. Colour is used to encode information in information visualizations. An example of such an encoding is categorization using colour (e.g., green for land, blue for water).<p> As a result of the variation in colour differentiation ability among individuals, many people experience difficulties when using colour-encoded information visualizations. These difficulties result from the inability to adequately differentiate between two colours, resulting in confusion, errors, frustration, and dissatisfaction.<p> If a user-specific model of colour differentiation was available, these difficulties could be predicted and corrected. Prediction and correction of these difficulties would reduce the amount of confusion, errors, frustration, and dissatisfaction experienced by users. This thesis presents a model of colour differentiation that is tuned to the abilities of a particular user. To construct this model, a series of judgement tasks are performed by the user. The data from these judgement tasks is used to calibrate a general colour differentiation model to the user. This calibrated model is used to construct a predictor. This predictor can then be used to make predictions about the user's ability to differentiate between two colours.<p> Two participant-based studies were used to evaluate this solution. The first study evaluated the basic approach used to model colour differentiation. The second study evaluated the accuracy of the predictor by comparing its performance to the performance of human participants. It was found that the predictor was as accurate as the human participants 86.3% of the time. Using such a predictor, the colour differentiation abilities of particular users can be accurately modeled.