In the process of infrared and visible image fusion, the details of the fusion image are often missing. In order to solve this problem, an improved image fusion algorithm based on nonsubsampled contourlet transform (NSCT) was adopted, which integrated the dynamic weighted non-negative matrix factorization (WNMF) rule into the image fusion. The results show that, the low-frequency and high-frequency parts can be obtained by multi-scale and multi-directional decomposition of two source images by using NSCT algorithm. The dynamic WNMF fusion rule was used as the fusion rule of low frequency parts. The decomposition scale of the highest level in the high frequency part was based on the absolute value method. The matching threshold was set in the other parts of the high frequency part. When the threshold value was lower, the method of region variance based on region energy matching degree was used. When the threshold value is higher, the weighted average method was used. By processing the low frequency part and the high frequency part, the fused image was obtained by NSCT inverse transform. This method effectively improves the definition of the fused image, highlights its details and shortens the calculation time.