The current mainstream distance metric learning approaches that all need to so lve the positive semi definite programming problem (SDP) will lead to high compu tational complexity, and they are thus difficult to be applied to large scale dat asets well because of fully matrix characteristics decomposition operational in each loop iteration. A distance metric learning method based on fe ature grouping and eigenvalue optimization is proposed considering the above pro blems. Firstly, a feature grouping algorithm is introduced to segment i mage features into several groups according to the correlations between each dim ension of characteristics. Then, the SDP problem can be covered to eigenvalue optimization issue under some certain constraints. Therefore, only the maximum eigenvalues of matrix is needed in every loop iteration. Experiment results indi cate that the computational complexity and the learning time of metric matrix ar e reduced effectively. Besides, the classification results are improved compared with the traditional methods.