Estimation de la réflectance de matériaux d'une scène urbaine: Modélisation et méthode d'inversion

Urban terrestrial images are widely available to web users, mainly in urban areas, through immersion services such as Google Street View. In order to enhance the visualization of these images and provide their information in a comparable way, radiometric corrections must be performed to free them from illumination conditions at the time of acquisition. The modelization of the radiance hitting the sensor must take into account both atmospheric scattering and objects intersections, which can be of high importance in urban images with high spatial resolution. In that work, the radiance is simulated by a ray-tracing algorithm, which is validated on simple scenes. Then, given the simultaneously acquired 3D geometric model of the scene with lidar or vision techniques, we face an inverse rendering problem where the illumination and the geometry of the scene are known and the reflectance of the scene is to be estimated. The proposed method uses a symbolic ray-tracing rendering to generate parametric images, for quick evaluation and comparison with the acquired images. The method is then based on an iterative estimation of the reflectance parameters of the materials, using a single symbolic rendering processing. The results on synthetic scenes are promising with linear BRDF models, and the limitations of the proposed approach with more general BRDF models that are non-linear with respect to their parameters are discussed.