利用空間分割及具深度貼圖的複雜場景顯像技術

Abstract

本論文主要是結合3D幾何式與2D影像式顯像技術的優點，並且利用空間分割、影像曲化等技術，發展出一顯像時間能與場景複雜度較無關的虛擬實境瀏覽系統。整個系統將分前置處理與執行兩階段，前置處理階段首先將整個場景作雙層蜂巢式的空間分割成為正六邊形的影像與瀏覽子空間，且每一瀏覽子空間對應一個影像子空間，而且有相同的中心點。再對每一個影像子空間，以其中心當作視點，六個邊面為視窗，分別產生六張快取影像，並且將深度值也一併儲存，最後將深度快取影像作狄勞尼三角化處理，以得到深度貼圖網格。執行階段則是瀏覽時整合影像子空間內幾何式顯像與影像子空間外影像式顯像的結果。對於空間外的場景利用深度快取影像與深度貼圖網格經過影像曲化與重新投影的運算，代替場景資料進行影像式顯像，而對於子空間內的幾何物體則選擇適當精細度的漸進式模型，配合可見性裁切決定出最後的幾何資料供顯像系統進行幾何式顯像。

In this thesis, we combine geometry-based and image-based rendering techniques to design and implement a VR navigation system that will have efficiency relatively independent of the scene complexity. The system has two phases. In the preprocessing phase, the x-y plane of a 3D scene is partitioned into equal-sized hexagonal cells, called navigation cells, each of which is associated with a larger image cell that has the identical center. Each side face of the image cell will be stored a cached image with depth that is obtained by rendering the scene using the cell's center as the projection center and the side face as the window. The depth mesh of the cached image will be obtained by triangulating cached image using depth. In the run-time phase, the participant navigates inside a navigation cell and views the image derived by combining the geometry-based rendering of the objects inside the corresponding image cell and image-based rendering of the objects outside the corresponding image cell. Objects outside the image cell will be rendered by warping and reprojecting depth mesh with cached texture image and objects inside the image cell will be rendered by using meshes with appropriate resolution. The visibility culling technique will be also integrated to speedup geometry-based rendering.