This paper presents a new model resizing approach intended for preservation of important geometric content and geometric symmetries of the model to be resized. We first extract high-level symmetric regions and other low-level salient regions from the input model. Then, we map the extracted low-level and high-level geometry information to a protective tetrahedral mesh around the model, and we define a symmetry-preserving and content-aware resizing function on the tetrahedral mesh using 3D mean value coordinates space deformation. By interpolating within the resized tetrahedral mesh, we obtain the final resizing results for the embedded model. Interactively defined user constraints are also incorporated into our framework to produce more desirable results. Our results show that the resized models preserve the geometric features of important regions in addition to the symmetric aspects of the original model, and eliminate the undesirable distortion in other less important regions.
- Geometry saliency
- Model resizing
- Protective mesh
- Space deformation
ASJC Scopus subject areas
- Computer Vision and Pattern Recognition
- Computer Graphics and Computer-Aided Design