具材質保留的觀視點相依多重精細度模型

Abstract

強調互動與融入的虛擬實境應用要求即時的快速顯像與逼真的成像品質， 觀視點相依多層次精細度技術希望對模型局部區域即時地改變精細程度， 讓顯像所處理的多邊形集中在可見範圍內， 具有隨觀視點改變迅速更新能力，在保留視覺效果同時兼顧到互動需求。 本論文在一套兼顧幾何特徵保留、拓樸架構調整，以及材質保 留功能的頂點叢集簡化演算法的根基上，研究出一套 物體模型局部精細程度隨觀視點移動而能快速調變之演算法。 我們提出以簡化步驟間相依關係建立相依圖形結構做為局部調變精細程度 之參考依據，並且區分相依關係為繼承關係與依賴關係，以便於 有效運用時間與空間之共效性， 而在動態瀏灠運作時，則儘可能地減少需要判斷與更新的資料數量。 實驗結果證實我們的演算法可以在各種觀視距離與觀視點移動情況下 ，仍然有穩健的精細度調變處理時間的表現。

Level of detail (LOD) modeling or mesh reduction has been found useful in interactive walkthrough applications. In particular, view-dependent LOD modeling has its strength in selective refinement and effective view and back-face culling. In this paper, we propopse a view-dependent LOD modeling in which the selective refinement is performed according to a dependency graph. The dependency graph is constructed based on a progressive meshing scheme that is clustering-based and takes both geometry and topology simplification into account. In the dependency graph, nodes represent the simplification steps while the edges representing relations between nodes. The relation associated with each edge is classified as either inheritance or dependence relation in order to facilitate the use of spatial and temporal coherence. In the run-time or selective refinement phase, a scheme is proposed to reduce the amount of selective and gauged nodes. Our experiments reveals that the proposed scheme provides a steady and efficient computational efficiency in the process of selective refinement.