Abstract: Image fusion is an important research field of image processing. How to get the best fusion quality is not an easy problem for the researchers. The nonsubsampled contourlet transform (NSCT) is a multi-resolution tool for image fusion. For NSCT, how to get a better focus measurement is an important research content. In this study, a new model named dual-layer PCNN model is proposed. It simulates human visual perception mechanism. The model not only takes into account local neighbor relativity each other but also takes into account the relativity between before and after layers. Compared with the other PCNN models, this model uses the Shannon information entropy to adaptively control its iteration process. Based on this model and NSCT, a new image fusion method is proposed. In the method, the source images are decomposed by NSCT firstly and then the dual-layer PCNN model and local energy match rule are used to select the coefficients. At last, the fused image is reconstructed by taking an inverse NSCT. The experimental results show that the dual-layer PCNN model is a good focus measurement for NSCT and the method proposed in this study has better fusion performance than the other classical methods.