Fast Face Recognition Algorithm Based on Collaborative Representation Classification and Manhattan Norm

We propose two simple and efficient algorithms to improve the efficiency of collaborative representation based classification with regularized least square (CRC_RLS) in Face Recognition. Based on the observation that the reconstruction coefficient vector infinite-norm is correlated to the corresponding test sample label, we propose CRC with infinite-norm (CRC_IN) algorithm, which greatly improve the speed. Furthermore, we improve the classification accuracy by proposing another algorithm. The CRC with Manhattan-norm (CRC_MN) algorithm is proposed by utilizing the strong correlation between the maximal Manhattan-norm of all groups in reconstruction coefficient vector and test sample label. The extensive experiments clearly show that CRC_MI run faster than CRC_RLS by about 47~131 times, the accuracy is the same or similar to CRC_RLS.