階層性以特徵為基礎的容積資料形變之研究

Abstract

形變技術可以被廣泛地使用在各種領域，然而二維影像形變技術，存在著難以克服的三維資訊失落的問題，使得三維形變技術的研究被熱烈地討論。在本篇論文內，我們應用三維容積資料的圖學技術，提出一種以階層式特徵為基礎的形變演算法。基於1992年Beier and Neely這二位學者所提出的以特徵為基礎的形變技術，我們提出一種新的特徵元件 – Cylinder Field，並且以階層性的架構來組織這些特徵元件，用來改進以特徵為基礎的形變演算法。應用我們的形變演算法，可以很平順地表現出從來源物體到目的物體的形變過程。除此之外，本論文中所提出的階層性特徵形變演算法亦可應用於容積資料的動畫。

Morphing techniques can be broadly applied to many fields. In two-dimensional image morphing, when source and destination images are rendered from 3D models, inaccurate morphing results may be generated, because 3D geometric information of original 3D models may be lost in source and destination images. Hence, three-dimensional morphing techniques are expected to have satisfactory solution. In this thesis, we propose a hierarchical feature-based volume morphing algorithm. We base on feature-based image morphing proposed by Beier and Neely [1] and propose a new type of feature element – Cylinder field to specify object features. Furthermore, we also propose a feature hierarchy to improve feature-based morphing. Using our method, we can create smooth and realistic morphing results. In addition, we propose a volume animation algorithm that leverage hierarchical feature elements.