結合三維定位與反光特性之表面最佳化技術

Abstract

本論文提出一個表面最佳化的演算法，當對真實的物體做重建動作且最佳化表面的同時亦可以取得其反射特性。我們使用結構光三維定位的系統去獲得一組較精確的起始三維模型。接著使用Phong和一個雙向次表面散射反射分配函數的模組來逼近真實場景的光反射特性。最後，在最佳化三維位置和光反射特性後，一組較精確的三維表面就可以被取得。不像大部份之前對於三維重建的方法，重建一些非只有散射特性或具次表面散射的材質物體是一個麻煩的問題。而且當我們重建了一組更精確的三維表面之餘，此物體的光反射特性我們亦取得，可以使用在描繪於虛擬場景時更真實且多變。

This thesis presents an optimization algorithm to simultaneously estimate both the 3D shape and parameters of a surface reflectance model from real objects. We use stereo structured light system to obtain initial stereo positions. And the Phong and the BSSRDF reflectance model are used to approximate the surface reflectance. After optimizing both of the shape geometry and reflectance properties, a more accurate surface can be acquired. Unlike most previous work in shape reconstruction, the proposed method deals with the troublesome problem of scanning Non-Lambertian and subsurface-scattering objects. In additional to a more accurate surface, the reflectance parameters can also be used for advanced rendering.