The invention discloses an image super-resolution method for an adversarial generative network based on fusion mutual information, and the method comprises the steps: carrying out the downsampling ofa high-resolution image in an image data set, generating a corresponding low-resolution image, packaging the high-resolution image and the low-resolution image into an h5py format in a one-to-one correspondence manner, and enabling the high-resolution image and the low-resolution image to serve as a training data set; using a low-resolution image and a corresponding high-resolution image in a training data set as input of a model; training a generation network, a discrimination network and a mutual information estimation network in the model by taking the corresponding high-resolution image asa target high-resolution image to obtain a deep neural network model capable of performing super-resolution on the low-resolution image to obtain a high-resolution image; and performing super-resolution processing on the low-resolution image in the test data set by using the trained model. Through the image super-resolution model of the confrontation generation network based on fusion mutual information, an image with a very good perception effect can be generated.