Sparse representation classification (SRC) and collaborative representation classification (eRe) are the most promising classifiers for classifying high dimensional data. However, they may suffer from outliers and noises, as l2-norm on signal fidelity is not effective enough to represent the test sa