Framework for Evaluating Visual and Geometric Quality of Three-Dimensional Models

Abstract

Three-dimensional (3-D) models require a quality evaluation prior to their use in specific applications, regardless of whether the modeling process is automatic or manual. A comprehensive framework to evaluate visual and geometric quality of 3-D models is proposed in this study for use in various applications. To evaluate the visual quality, we propose a new concept that we call a fuzzy realistic index, which is based on fuzzy set theory and information entropy methods. Geometric accuracy is determined by deconstructing a 3-D model into the elements of point, line, and plane. The campus of Nanjing University, China, is the experimental area for this study, consisting of 106 3-D building models having various styles and complexity. In this experiment of visual quality, we selected 30 models and invited 100 participants in order to illustrate the power of the proposed fuzzy realistic index. Furthermore, with the support of the proposed framework, this study analyzes the influence of building complexity and of the perceptions of different people upon the quality of constructed 3-D models. Results of these experiments reveal the applicability of the proposed framework in evaluating the quality of 3-D models.