Collaborative representation-based classification has shown promising results on cognitive vision tasks like face recognition. It solves a linear problem with l1l1 or l2l2 norm regularization to obtain a stable sparse representation. Previous studies