Abstract. There has been great success in speeding up global illumination computation in diffuse environments. The concept of clustering allows radiosity computations even for scenes of high complexity. However, for lighting simulations in complex non-diffuse scenes, Monte-Carlo sampling methods are currently the first choice, because non-diffuse finite-element approaches still exhibit enormous computation times and are thus only applicable to scenes of very modest complexity. In this paper we present a novel clustering approach for radiance computations, by which we overcome some of the problems of previous methods. The algorithm computes a radiance solution within a line space hierarchy, that allows us to efficiently represent light propagation and reflection between arbitrary non-diffuse surfaces and clusters.