Key node recognition is an essential component in the field of complex networks. To address the problems of low resolution and high computational cost of key node recognition methods in complex networks, an improved mixed degree decomposition method (IMDD) is proposed in this paper to identify key nodes. Firstly, The mixed degree decomposition method is used to stratify the network and determine the mixed K-shell (Km) value of each node to globally quantify the importance of the nodes. Secondly, the concept of the influence coefficient of sub-neighbors is proposed, and the dynamically adjustable influence coefficient of sub-neighbors of each node is calculated. Finally, a formula for the comprehensive score of node importance is constructed, and the comprehensive score of each node is calculated by balancing the influence of neighbors and sub-neighbors and thus ranking the importance of nodes in the network. The parameters in the Km calculation formula are pre-tested in six realistic networks of different sizes and compared with the other six classical methods through SIR contagion model simulation experiments. The experimental results show that changes in parameter values have minimal impact on the identification of key nodes, and the method can effectively identify key nodes, outperforming the other six classical methods in terms of resolution, accuracy, and relevance, and maintaining a lower The results show that the method can effectively identify key nodes and outperforms the other six classical methods in terms of resolution, accuracy and relevance, and maintains a low time complexity.