Aiming at the problem of insufficient feature information extraction during the fusion course of visible light and infrared images, the fusion technique for visible light and infrared images based on non-subsampled shearlet transform (NSST) and deep Boltzmann machine (DBM) is proposed in this paper. It obtains the significant infrared targets after having optimal energy segmentation by making use of the deep Boltzmann machine (DBM), as well as puts the mapping graph combined the infrared targets region with the background region into sparse decomposition and fusion by adopting the non-downsampled shear wave transform (NSST) . The simulation results show that the results based on the proposed method have better visual effect and better objective performance compared with the existing classical methods.