In order to improve the visual effect of infrared and visible light image fusion and overcome the artifact effect of the fusion result, an image fusion method based on internal generative mechanism and convolution sparse representation is proposed. First, the source image is decomposed by internal generative mechanism that conforms to the reasoning of the human brain to obtain the prediction layer and the detail layer. Then the detail layer is decomposed using a convolution sparse representation to obtain the secondary detail layer and the basic layer, and adopts the activity level measurement to be larger and the weighted average rule for fusions separately; defines the ISR hybrid operator fusion rule for the prediction layer and fuse; finally, the fusion prediction layer and detail layer are added to obtain the final fusion result. In the experiment, this paper uses three representative infrared and visible light images for algorithm testing. The experiment shows that the proposed method has good subjective visual effects, and the objective evaluation indicators are also better and effective.□□□□□□□□□□□□□