Two-stage rain and fog image restoration method based on conditional diffusion model

The invention discloses a two-stage rain and fog image restoration method based on a conditional diffusion model, and belongs to the technical field of digital image processing, and the method comprises the following steps: 1, obtaining rain and fog image data to be restored, and carrying out the preprocessing of the rain and fog image data; 2, inputting the processed image into a trained degradation perception model for processing, and removing a weather degradation foreground irrelevant to the background content of the image to obtain an output X of a first stage; 3, inputting the output X in the step 2 into a semantic information extractor to obtain a semantic feature F; and 4, simultaneously sending the output X in the step 2 and the output F in the step 3 into a semantic perception condition diffusion model to obtain a final restored image. According to the method, a rain and fog image restoration task is decomposed into two sub-tasks of degradation removal and content reconstruction, texture details whi