This paper presents a novel approach for automatic target recognition (ATR) using inverse synthetic aperture radar (ISAR) images. This proposed approach is mainly com- posed of two steps. In the rst step, we adopt a statisti- cal method to compute a novel target template from fea- ture descriptors. The proposed template is achieved by combining the Gamma statistical parameters of the both dual-tree complex wavelet transform (DT-CWT) coecients and the scale-invariant feature transform (SIFT) descrip- tor. In order to validate the proposed target template, we achieve in the second step the recognition task using a sparse representation-based classication (SRC) method. The performance of the proposed approach has been success- fully veried using ISAR images reconstructed from anechoic chamber. The experimental results show that the proposed method can achieve a high average accuracy and is signi- cantly superior to the well-known SVM classier.