歐陽明臺灣大學：資訊工程學研究所吳宗益Wu, Tsung-YiTsung-YiWu2007-11-262018-07-052007-11-262018-07-052006http://ntur.lib.ntu.edu.tw//handle/246246/53875In order to create photo-realistic images, we need to know how light is reflected from different materials. The bidirectional reflectance distribution functions (BRDF) plays an important role in modeling reflectance of materials. In this thesis, a process is presented to extract per-pixel BRDF measurements from multiple photographs of an object from the same viewpoint but under different directions of light sources. After the measurement is done, we use a statistical smoothing method to estimate the whole BRDF from measured data.1 Introduction 5 1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.1.1 BRDF . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.1.2 Classes of BRDFs . . . . . . . . . . . . . . . . . . . . . 8 1.1.3 Properties of BRDFs . . . . . . . . . . . . . . . . . . . 9 1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.3 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . 11 2 Previous Work 12 2.1 BRDF Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.1.1 Phenomenological models . . . . . . . . . . . . . . . . 13 2.1.2 Physical-based models . . . . . . . . . . . . . . . . . . 13 2.2 BRDF Measurement . . . . . . . . . . . . . . . . . . . . . . . 14 3 Data Acquisition 16 3.1 Light Calibration . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.2 Acquiring Images . . . . . . . . . . . . . . . . . . . . . . . . . 18 3.3 Computing Surface Orientation . . . . . . . . . . . . . . . . . 20 3.3.1 Lambertian photometric stereo . . . . . . . . . . . . . 21 3.3.2 Outlier removing . . . . . . . . . . . . . . . . . . . . . 22 3.4 Segmentation for Objects with Multiple Materials . . . . . . . 26 3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 4 Data Presentation and Interpolation 28 4.1 Coordinate System . . . . . . . . . . . . . . . . . . . . . . . . 29 4.2 Interpolation Scheme . . . . . . . . . . . . . . . . . . . . . . . 32 4.2.1 The Epanechnikov kernel . . . . . . . . . . . . . . . . . 32 4.2.2 Searching strategy . . . . . . . . . . . . . . . . . . . . 32 5 Result 35 6 Conclusions and Future Work 43 6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 6.2.1 Precise surface normal estimation . . . . . . . . . . . . 44 6.2.2 BRDF measurement for objects with complex shape . . 44 6.2.3 Reflectance representation . . . . . . . . . . . . . . . . 456416515 bytesapplication/pdfen-US擷取雙向反射分佈函數影像方式AcquisitionBRDFImage-basedthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/53875/1/ntu-95-R92922056-1.pdf