Abstract
摘 要 近幾年來, 因電腦計算能力的大幅提升, 使得3D即時成像(real-time rendering)的技術日益成熟. 許多以電腦畫出的3D場景, 不但其逼真的程度可與相片媲美. 而且更能以動態遊走的方式, 達到虛擬實境互動的目的. 然而虛擬場景的精細程度可以無限制的要求, 但是儘管硬體的進步如何, 其運算的速度總是有限, 因此在塗彩(shading)時, 大量的三角片數雖然可以達到逼真的效果, 但是往往會造成系統的負擔而無法即時顯像. 因此如何簡化模型中的三角片數, 而盡量保持圖像的逼真程度, 是計算機圖學中相當重要的研究課題. 在計算3D成像的領域中,廣域照明(Global illumination)的運算能真實地模擬出虛擬世界中光影成像的效果. 其中如輻射成像法(radiosity method), 利用環境中能量交換的公式, 模擬出逼真的圖像. 然而經過輻射成像法的計算, 一個3D立體場景的幾何複雜度(geometric complexity)因精度的要求而大幅增加, 所以如果要即時將這些龐大的資料塗彩, 會需要佔用到大量的系統資源, 影響系統的速度. 本研究將提供新的方法, 可以整合因輻射成像法所產生出過於細碎的多餘的三角片; 而仍然保持了場景中光影的重要特徵以及精細的品質. 除此之外, 將此一靜態的簡化結果當做前置作業之後,我們也提供了一個能夠動態自動產生多精度模型(LOD model)的演算法,依照使用者觀察的位置決定3D場景的精細程度. 使系統在更有效率地應用資源同時, 也能保持模型成像的品質, 達到真正即時互動的目標.Abstract Recently the advance of computer technology hasenabled the sophisticated techniques of image synthesis toproduce photorealistic images. The generation of realisticimages requires a precise model of lighting effect whichsimulatesthe physical phenomenon of light emission propagationand reflection. The radiosity method is one of these lightingmodels which is account for the global illumination. It wasfirst introduced to computer graphics communityto deal with thelighting effect of diffuse environment. However, the algorithmtoapproximate of energy distribution function usually requiresvery fine subdivision of the scene geometry. The excess amountof subdivided polygons generated after the computation of energyexchange will slow down the renderingprocess. In thisstudy, we will propose an algorithm to simplify the excessivemeshes due to radiosity computation at small expense of imagequalities. The basic idea of our approach is to regard theapproximation of energy distribution function as a contour meshover the underlying geometry. Then the polygon meshes of colorintensities are simplified within given visual tolerance and theimage quality is assured. Moreover, the simplified intermediatedata can be used as level of details for the application of realtime walkthrough.