Towards Effective Measurement and Interpolation of Bidirectional Texture Functions

Bidirectional texture function (BTF) is acquired by taking thousands of material surface images for different illumination and viewing directions. This function, provided it is measured accurately, is typically exploited for visualization of material appearance in visual accuracy demanding applications. However, accurate measurement of the BTF is time and resources demanding task. While the sampling of illumination and viewing directions is in all known measurement systems done uniformly, we believe that to be more effective the sampling should be tailored specifically to reflectance properties of materials to be measured. Hence, we introduce a novel method of sparse BTF sampling. The method starts with collecting information about material visual behavior by means of small initial subset of reflectance samples measurement and analysis. This information is fed into our heuristic algorithm producing sparse material dependent sampling that is consequently used for BTF measurement and interpolation. The algorithm was tested in simulated measurement test with ten BTF samples, their estimated image subsets were selected, the remaining images were interpolated, and results were computationally and psychophysically compared with the measured data. In average the number of sampling points was less Than half the number of original measurements points, and for most materials the produced BTF renderings were perceptually indiscernible from the originals.