A texture-mapping scheme for volumetric models

Abstract

In this article, we propose a texture-mapping procedure to enhance visual appearances of volumetric models. To texture a volumetric model, the proposed method performs an unsupervised training to transform a Self-Organizing Map (SOM) into a fitting surface at first. Then, another unsupervised training is conducted to UV-parameterize the fitting surface. As a result, a bridge connecting the volumetric model's surface and the texture map has been established. In the following steps, filtering computations are performed in the fitting surface and the texture map to compute texel values for the model's surface voxels and to reduce aliasing artifacts. Experiments confirm that the proposed procedure is able to efficiently and smoothly texture volumetric models within acceptable time. The test results also reveal that the quality of the texturing process is influenced by the filtering algorithms and the topology and resolution of the embedded SOM. Further trials indicate that texturing results can be improved by using image warping, affine transformations, anchor points, and region-of-interest selection.