Estimating Reflectance Properties Based on Semi-Parametric Model

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K00239
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Perceptual information processing
• Research Institution: Kyushu University
• Principal Investigator: Hara Kenji 九州大学, 芸術工学研究院, 准教授 (50380712)
• Co-Investigator(Kenkyū-buntansha): 浦濱 喜一 九州大学, 芸術工学研究院, 教授 (10150492) ; 井上 光平 九州大学, 芸術工学研究院, 准教授 (70325570)
• Co-Investigator(Renkei-kenkyūsha): URAHAMA KIICHI 九州大学, 大学院芸術工学研究院, 教授 (10150492)
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: 異方性BRDF / 変数分離 / 最大事後確率推定

Outline of Final Research Achievements

We developed a new method for reconstructing real-world BRDFs from BRDF data, which are raw data captured by exhaustively tabulating reflectance from many sampled incident and reflected directions. More specifically, we derive a low-dimensional semi-parametric model to perform interpolation directly in the parameter space to the missing reflectance data. The approach employed in this study is as follows. The microfacet distribution in an anisotropic reflectance model is first approximated as a product of two constrained mixtures of Gaussian distribution functions. Then, this approximation leads fitting low-dimensional parametric models to the acquired reflectance data to provide a maximum a posteriori (MAP) estimation. Finally, unknown set of model parameters is determined by minimizing Kullback-Leibler distance between expected and observed probability density functions. The reconstructed BRDF is then used for completely rendering virtual objects from missing BRDF data.

Free Research Field

コンピュータビジョン・画像処理