Self-supervised smoke removal method for operation video

The invention discloses a self-supervised smoke removal method for an operation video, and belongs to the technical field of image restoration. The problem that operation smoke generated in the laparoscopic operation process disturbs the visual field of a doctor is solved. A video smoke removal model is built by a smoke frame encoder, a reference frame encoder, an alignment module, a fusion module and a reconstruction module, a frame before a high-energy instrument starts to work, namely a pre-smoke frame, is regarded as a non-alignment supervision signal, a smoke video sequence # imgabs0 # containing N smoke frames and the pre-smoke frame Sref are adopted to train the video smoke removal model, and the non-alignment supervision signal is obtained. Obtaining a trained video smoke removal model; in practical application, the trained video smoke removal model obtains a smoke-removed video sequence based on a real pre-smoke frame and a real smoke video sequence. The method is used for removing the smoke in the s