Using directional statistics to learn cast shadows from a multi-spectral light sources physical model

摘要: In this paper is proposed a novel statistical learning approach, to identify cast shadows, and model their generation. We exploit the theoretically well-founded directional statistics field, in order formulate generation of shadows as Mixture Von Mises-Fisher distributions (MovMF) on unit sphere. This formulation based bi-illuminant physical where no prior assumptions spectral power distribution (SPD) direct light sources ambient illumination scene are made. Founded rigorous parametric framework capable modelling shaded surface behavior complex scenes meet real time requirements. better discriminating provides more compact representation, achieve accuracy, with less data much computation time, compared non-parametric models previously proposed. Theoretic analysis experimental evaluations demonstrate effectiveness