This paper proposes a dimensionality reduction method for hyperspectral data based on manifold regularization nonnegative matrix factoriza tion (MR-NMF).The new method describe s the data geometry by constructing the neighbor graph of the sample points,and then adds it as a regular term to the objective function of the NMF for combinatorial optimization . Experiments were done in HYDICE dataset.The experimental results show that the new method can improve the accuracy of hyperspectral image classification.