Retrieval of Colored 3D Models

The number of 3D models that are available from the Web is increasingly growing, motivating the research on content-based retrieval from shape information. In this paper we propose to extend the approach to colored 3D models, which start to become more and more popular. To this end we propose a content descriptor that combines together shape and color information and it is based on an extension of spin-images, where also texture information is taken into account. The proposed approach has been evaluated on a collection of 600 automatically colored 3D models, considering as relevant to a query only the models with the same shape and the same added texture. Results show that the approach gave good results in terms of retrieval effectiveness, also if compared to a pure shape-based approach. Issues related to computational cost have been addressed proposing two alternative improvements that have been evaluated and compared with the basic approach.