Personalized Retrieval from Material Color Databases for 3D Polygonal Models

There has been considerable demand for the development of virtual spaces. When virtual spaces such as virtual cities or virtual offices are created, the material colors of 3D models in these spaces greatly influence the appearance, so designers of virtual spaces must use proper material colors. Material colors of 3D models can be specified with ambient color, diffuse color, specular color, emissive color, shininess and transparency. These values can be changed interactively by using a graphical user interface based on material editors, and the user can see the effects of those changes. Once the desired material colors have been created, users are required to store them into a personalized material color database. Since users often use subjective terms to describe material colors, it would be useful if the user could retrieve material colors by using subjective terms such as ‘romantic’ and ‘warm.’ Since commercially available material color databases contain huge amounts of data, manually described subjective terms for each color are not possible, therefore, computerized indexing techniques are needed. We have analyzed the correlation between subjective term and databases of material colors by using statistical methods so that retrieval can meet each user’s preference model. The statistical methods include principal component analysis and multiple regression analysis. We have developed a subjective terms indexing technique to retrieve material colors from a database for 3D models by using the results of these statistical analyses. The technique enables users to retrieve the subjective terms of material colors reflecting users’ personal preferences. We have also developed visualization tools to determine how users are using subjective terms to describe material colors.