Interactive Volume Deformation Based on Model Fitting Lattices

Volume visualization, which is a relatively new branch in scientific visualization, not only displays surface features of a model, but enables an intuitive presentation of the internal information of the object. Its comprehensive visualization algorithms developed in the last decade have brought in challenges such as complex data processing, real-time operations, and application-specific system performances. These challenges were elaborated in the manner of research objectives in the thesis.

By devising a novel volume deformation pipeline, this thesis managed to explore volume-model related operations applied for complicated applications through illustrating the feasibility of the designed system that was verified by experimental results. The contribution of the programme was demonstrated via testifying the effectivities of the four system design characteristics. Firstly, the clustering-based segmentation methods were adopted by the volumetric data processing module within the proposed volume deformation system for managing the complicated structures often existing in large volume data sets. Secondly, a novel mesh construction method was formulated in terms of optimizing the control lattices for the following deformation process. Thirdly, the volume deformation approach devised in the research has taken advantages of the parameterization process of the entire shape-change process. Finally, the GPU-based parallel process architecture was utilized to accelerate the calculation of Gaussian sampling in the lattice construction process; the progressive locations of the removed points in the simplification scheme; and the integration of kinetic energy for determining the deformation behaviours.