Abstract: In this paper, we consider the problem of categorizing videos of dynamic textures under varying view-point. We propose to model each video with a collection of linear dynamics systems (LDSs) ...