Wojciech Matusik presents a data-driven model for isotropic bidirectional reflectance distribution functions (BRDFs) based on acquired reflectance data. Instead of using analytic reflectance models, each BRDF is represented as a dense set of measurements, allowing for interpolation and extrapolation to create new BRDFs. The acquired BRDFs are treated as high-dimensional vectors, and both linear and non-linear dimensionality reduction techniques are applied to discover a lower-dimensional representation that characterizes the BRDFs. Users can define perceptually meaningful parametrizations to navigate in the reduced-dimensional BRDF space. The model also derives efficient BRDF sampling procedures that require fewer measurements than standard uniform sampling approaches. By analyzing a large collection of reflectance data, Matusik derives two novel reflectance sampling procedures that require fewer total measurements. These procedures use wavelet analysis to determine the optimal sampling pattern and basis functions for representing BRDFs. The model is validated through various tests, demonstrating its effectiveness in generating realistic BRDFs and handling complex effects like rust, oxidation, and dust. The thesis concludes with a discussion on future work, including the analysis of other surface reflectance functions, real-time rendering, and inverse methods.Wojciech Matusik presents a data-driven model for isotropic bidirectional reflectance distribution functions (BRDFs) based on acquired reflectance data. Instead of using analytic reflectance models, each BRDF is represented as a dense set of measurements, allowing for interpolation and extrapolation to create new BRDFs. The acquired BRDFs are treated as high-dimensional vectors, and both linear and non-linear dimensionality reduction techniques are applied to discover a lower-dimensional representation that characterizes the BRDFs. Users can define perceptually meaningful parametrizations to navigate in the reduced-dimensional BRDF space. The model also derives efficient BRDF sampling procedures that require fewer measurements than standard uniform sampling approaches. By analyzing a large collection of reflectance data, Matusik derives two novel reflectance sampling procedures that require fewer total measurements. These procedures use wavelet analysis to determine the optimal sampling pattern and basis functions for representing BRDFs. The model is validated through various tests, demonstrating its effectiveness in generating realistic BRDFs and handling complex effects like rust, oxidation, and dust. The thesis concludes with a discussion on future work, including the analysis of other surface reflectance functions, real-time rendering, and inverse methods.