To achieve the goal of photorealism in computer graphics means better scene geometry modeling better global rendering algorithms and better local re ection models These models are represented by BRDFs bidirectional re ectance distribution functions We study in this thesis some analytical BRDFs models and compare them with measured BRDFs data This study leads to the conclusion that real world diversity cannot be handled by a uni ed BRDF model and that it is necessary to directly use measured BRDFs data in image computation We then present existing methods to represent these data and we introduce a new method particularly tted to BRDFs special case This method is based on locally supported functions and a hierarchical approach In order to validate our BRDFs reconstructions we introduce also an error criterion so as to compare BRDFs To create a link between measured BRDFs and computed BRDFs we present a software that simulates the behavior of a real goniophotometer a device used to measure BRDFs This virtual goniophotometer allows us to validate BRDFs analytical models and to compare real BRDFs and the result of a simulation from measured material properties
