結合快取影像式之網路顯像技術

Abstract

本論文中將描述一套在網路環境下，以影像快取式顯像為主體的顯像方法。這個方法可以達成一個比較穩定的顯示頻率，不會因為場景複雜度之提高而造成顯示計算的大量增加，並且達成在網路上傳送場景資料的時間與整個場景複雜度無太大關係的目的。架構上是以主從式建構出一個網路顯像模式，將場景以蜂巢式之分割方式，在 x-z 平面上劃分成許多大小相同之正六邊形的瀏覽子空間。對應每個瀏覽子空間並建構一相同中心點但大小比例大數倍的正六邊型當做影像子空間，以保證用戶端在瀏覽子空間內以影像子空間後的快取影像來顯像時，品質可以維持在一定的誤差內。場景分割之後，便將所有影像子空間的相關快取影像皆事先計算完成並且儲存於伺服端的大量儲存裝置 (如硬碟或光碟片)中。用戶端就根據所在位置相對應的子空間，經由背景傳送以及漸進傳送的方式由伺服端取得子空間邊界上經過壓縮的快取影像後，配合影像子空間內的幾何資料進行結合幾何式與影像式的顯像。

This paper presents an image-caching based rendering system for networked virtual reality applications. The proposed method aims to obtain a high and constant frame rate by reducing the rendering time and network transmission time due to the increase of scene complexity. The system is a client-server mode in which serever is responsible for providing scene data to the clients while each client renders the scent for a particular view. In the preprocessing performed in the server, we subdivide the scene horizontally into equal-sized hexagonal cells, which are called mavigation cells, and then for each navigation cell we construct a so called image cell, which is also a hexagon that has the same center as and the edges parallel to the corresponding navigation cell. The scene image behind each wall of each image cell is then rendered in the server and texture-mapped at the wall. All the image cells are a few times larger than the navigation cells such that, while navigating within a navigation cell, the quality of image rendered using the images cached at the walls of the corresponding image cell will be acceptable. For a particular view at a navigation cell, the client renders the scene by composing the cached images stored in the corresponding image cell and the image obtained by rendering the geometry objects inside the image cell. The data transmission between the server and a client is performed in background and in a progressive manner.