Dynamical Textures Modeling via Joint Video Dictionary Learning