Discriminative dictionary design using LVQ for hyperspectral image classification

In this paper, we propose a new technique for discriminative dictionary learning for hyperspectral image classification. The proposed algorithm generalizes the learning vector quantization scheme for sparse representation-based classifiers. It is known that a pixel can be represented by a sparse linear combination of atoms in a dictionary and its sparse representation vector contains the class information. The proposed learning technique utilizes the discriminative nature of the sparse vectors in the dictionary updating stage, generating a dictionary with both reconstructive and discriminative capabilities. Experimental results on a real hyperspectral data set demonstrate that using dictionaries learned from the proposed technique improves classification performance in various conditions.