Medical image registration method based on grouping deformation

The invention relates to the technical field of artificial intelligence, deep learning and computer vision content, in particular to a medical image registration method based on grouping deformation. Extracting the features of the moving image and the fixed image by using two weight sharing encoders; constructing a grouping module, extracting features of different receptive fields through a convolutional layer, and calculating feature similarity to generate a correlation graph; a context fusion module is constructed, the collaboration among different groups is enhanced, and the prediction accuracy is improved; and constructing a loss function which comprises regularization loss and weak supervision loss, ensuring continuity and smoothness of a deformation field, and guiding alignment of a network concerned region of interest. The invention aims to solve the problems of image dislocation and overlarge calculation amount when the existing pyramid registration algorithm is used for processing medical images. The