Traditional infrared and visible image fusion method decomposes images into several frequency components, fuses them separately, and then adds them together, resulting in problems of edge fuzziness, low contrast, and so on. The paper proposes a fusion method based on Tikhonov regularization and detail reconstruction. Firstly, images are decomposed into base layers and detail layers by Tikhonov regularization. A generative adversarial network is trained aiming at detail information reconstruction for base layers. Secondly, features of base layers to be fused are extracted, and the principal component analysis method is used for feature fusion. Finally, the fused results of base layers are input into generative network to reconstruct a fusion image with abundant high frequency information. Experimental results show that the method proposed in this paper preserves detail information and highlight areas of the source images well, with a good robustness to the images with different resolutions.