Non-linear reflectance model for bidirectional texture function synthesis

A rough texture modelling involves a huge image data-set - the bidirectional texture function (BTF). This 6-dimensional function depends on planar texture coordinates as well as on view and illumination angles. We propose a new non-linear reflectance model, based on a Lafortune reflectance model improvement, which restores all BTF database images independently for each view position and herewith significantly reduces stored BTF data size. The extension consists in introducing several spectral parameters for each BTF image which are linearly estimated in the second estimation step according to the original data. The model parameters are computed for every surface reflectance field contained in the original BFT data. This technique allows BTF data compression by the ratio 1:15 while the synthesised images are almost indiscernible from the originals. The method is universal, and easily implementable in a graphical hardware for the purpose of real-time BTF rendering.