In order to improve the accuracy and robustness of image matching algorithms based on spectral feature, a spectral feature matching algorithm based on maximum pool is proposed. Firstly, we use the neighborhood information of image feature points to extract spectral features that are invariant to rotation and the linearity of brightness changes. Secondly, we use the feature points described by spectral features as nodes and the Euclidean distance between feature points as edge properties, and transform the image matching problem into graph matching problem. Finally, the maximum pool matching strategy is introduced to obtain the graph matching results. A large number of experimental results show that the proposed algorithm improves the accuracy and robustness of the spectral feature matching algorithm.