Two kinds of hyperspectral image classification algorithms based on principal component analysis and local binary patterns are proposed. The principal component analysis is employed to reduce the redundant information in spectral domain. Following that, the local binary patterns are studied to analyze the spatial texture features. And the sparse presentation classification and support vector machine are used for a classification of extracted results, respectively. Combining the principal component analysis with the local binary patterns for extracting the features of hyperspectral image, we ensure that the spectral redundant information is reduced effectively, and the spatial local neighborhood information is protected. Hence, the proposed algorithms can not only sufficiently excavate spectral-spatial features of hyperspectral image for improving classification accuracy and Kappa coefficient, but also have outstanding classification performance in Gaussian noise environments and small-sample-size condition.