Self-supervised and self-adaptive image restoration and enhancement method based on surround photography

The invention discloses a self-supervised and self-adaptive image restoration and enhancement method based on surround photography, and belongs to the technical field of underlying visual reconstruction. The method aims at solving the problem that an existing image reconstruction model can only complete a single reconstruction task and is poor in real data reconstruction capability. Comprising the steps of obtaining a simulation data set and a real data set; constructing a time modulation cycle network composed of an encoder, an optical flow alignment module, a fusion module and a reconstruction module, pre-training the time modulation cycle network on a simulation data set, and performing fine tuning training on a real data set; performing normalization and gamma transformation on the T-frame multi-exposure RAW image to obtain a transformed image, performing feature extraction through an encoder to obtain encoding features, and performing alignment through an optical flow alignment module to obtain aligned f