Scene recognition remains one of the most challenging prob- lems in image understanding. With the help of fully con- nected layers (FCL) and rectified linear units (ReLu), deep networks can extract the moderately sparse and discrimi- native feature representation required for scene recognition. However, few methods consider exploiting a sparsity model for learning the feature representation in order to provide en- hanced discriminative capability. In this paper, we replace the conventional FCL and ReLu with a new dictionary learn- ing layer, that is composed of a finite number of recurrent units to simultaneously enhance the sparse representation and discriminative abilities of features via the determination of optimal dictionaries. In addition, with the help of the struc- ture of the dictionary, we propose a new label discrimina- tive regressor to boost the discrimination ability. We also pro- pose new constraints to prevent overfitting by incorporating the advantage of the Mahalanobis and Euclidean distances to balance the recognition accuracy and generalization per- formance. Our proposed approach is evaluated using various scene datasets and shows superior performance to many state- of-the-art approaches.