07-07-2023 дата публикации
Номер: CN116402706A
Принадлежит:
The invention discloses an image restoration method for a mutual coding and decoding network based on double-gating convolution, and the method comprises the following steps: S1, building a network model which comprises a restoration network and a discriminator network, and constructing a BG-MED network as the restoration network; s2, selecting a data set; s3, inputting the damaged picture into the BG-MED network model, repairing, and outputting to obtain a repaired picture; s4, calculating the total loss value of the discriminator, and updating the parameters of the global discriminator and the local discriminator by using an Adam optimization algorithm; s5, calculating a loss value of the repaired network, and updating parameters of the BG-MED model by using an Adam optimization algorithm; s6, repeating the steps S3-S5 for multiple times, and setting a network model value by using the trained model parameters; and S7, applying the BG-MED model to actual image restoration. The invention ...
Подробнее