基於多重BRDFs之複雜材質反射特性重建系統分析

李正倫
Rendering technology has many applications in diverse fields, for example, digital archives, multimedia entertainment, inverse engineering, and even medical science. Beside the acquirement of accurate 3D geometry, how to get the best fitting feature to reconstruct the lighting reflectance of the object from different viewing direction is a critical problem.

In the thesis, we use geometry-based rendering and BRFD model to represent the lighting response of materials. The purpose of our research is to build an automatically reflectance reconstruction system, which can automatically select the most suitable BRDF model from the database to describe the input object. And the selection process operates in accordance with the characteristics of the captured pictures from different viewing directions. The system embeds five most popular BRDF models in the database, i.e. Phong, Blinn-Phong, Cook-Torrance, Ward, and Ashikhmin model. Furthermore, in order to reconstruct the reflectance characteristic of complicated materials as well, we propose an effective cluster algorithm to compute the best fitting BRDF parameters. Most of the previous approaches use monochromatic or single characteristic material, and few researches for complicated materials. The proposed algorithm uses error vector to deal with the complicated materials in terms of the process of fitting, clustering, merging and reclustering. Thus, it could make cluster process robustly and better than the previous approaches for representing real-world objects.In the experiments, we apply the algorithm to several objects with different shape and complicated material to verity the effectiveness. The objects consist of different materials with varying reflection properties in both the diffuse and the specular part. Reflectance field is proved to be realistically reproduced from the object of complicated materials.

描繪技術(rendering technology)在不同的領域有很多普遍而重要的應用，例如數位典藏、多媒體娛樂、 工業製品、甚至醫療科學。除了擷取3D幾何外，如何從不同視角的影像中找出物體表面最佳的特徵，進而重建出物體表面的反射特性是個相當重要的課題。我們採用基於幾何之描繪技術，並使用BRDF Model來表達材質對光源的響應影響。本論文主要的研究目的是在建構一個自動化3D物體表面重建系統，此系統能依照相機從不同視角擷取下來的平面影像，自動計算並由系統中選擇出最合適的BRDF模型來描述並呈現此3D物體，其中包含五個重要的BRDF模型，如Phong、Blinn-Phong、Cook-Torrance、Ward和Ashikhmin模型。另外，為進一步處理複雜表面特性的物體，我們提出了一個新的群集演算法來計算最佳的BRDF參數。以往的研究大部分使用單色或單一種材質特性的物體，很少研究做到多種材質，相較於之前的研究者提出的方法，此演算法以誤差向量為條件，透過參數的近似、分群、合併、和再重新分群的程序，構成了強健的分群過程，更能處理複雜表面材質的物體。分群避免掉相同材質的頂點造成雪花雜訊的情形，增加不同材質特性分群的正確性，合併減少過多不必要的群數，使群數依據材質特性達到最小值，重新分群成左滷N分錯群的頂點回歸屬於自己的群體中，讓多種材質物體的反射特性重建得以實現。另外在參數計算的過程方面，演算法將漫射和反射部分分開計算，減少反射項的誤差對漫射項的干擾，避免整體誤差小，頂點誤差大的遺憾。實驗中，我們將由例子說明此自動化選擇系統的應用及其結果的正確性；此外，此演算法亦將應用於數個不同漫射和反射特性的材質物體上，經過參數的近似、分群、合併和重新分群的過程，重建出不同材質的反射特性，由結果的誤差量測及視覺上的呈現來驗證此演算法在處理複雜材質物體的能力。