Prediction of Visibility for Color Scheme with a Multi-stage Color Model

It is necessary to design web pages that users can access the information easily. In website design, color scheme is one of important elements. The prediction method of visibility of color scheme has been proposed. In this method, neural networks are proposed to predict paired comparison tables which show visibility of background and character colors. We improve this method and propose neural networks with a multi-stage color model, which is a model of human color vision. We propose how to extract color features with a multi-stage color model and to learn neural networks to predict visibility of color scheme. We collect data of visibility with paired comparison tests on the web browser and evaluate proposed method. The results obtained from our experiments show that proposed method can improve visibility of color scheme.