Distortion correction method based on neural network

The invention discloses a distortion simulation and correction method for extracting an aberration coefficient based on deep learning. The method comprises the following steps: 1) calculating a distortion aberration coefficient in a Sidel polynomial of a wavefront difference of a simulated image carrying distortion, calculating a wavefront difference function W, calculating a corresponding point spread function, and performing point-by-point convolution operation on the point spread function and the clear image to simulate an image carrying distortion, 2) taking the simulated image carrying distortion and the distortion aberration coefficient corresponding to the image as training pairs, inputting the training pairs into a convolutional neural network for learning training, and constructing a mapping relationship between the image carrying distortion and the distortion aberration coefficient, and obtaining a trained neural network model carrying the mapping relationship between the distorted image and the dis