A new sparsity-based classification algorithm for hyperspectral imagery is proposed in this paper. This algorithm is based on the assumption that the spectral signatures of pixels in the same class lie in a low-dimensional subspace and thus a test sample can be represented by a sparse linear combination of the training samples. The sparse representation is recovered by solving a constrained optimization and it directly determines the class label of the test sample. In addition to the constraints on sparsity and reconstruction accuracy, the smoothness of hyperspectral images across neighboring pixels is also exploited by forcing the Laplacian of the reconstructed image to be minimum in the optimization process. Various sparse recovery techniques are applied to solve the optimization problem and their performances are compared against the widely used Support Vector Machine classifier. Simulation results show that the proposed algorithm yields a favorable performance over the support vector machines.