Image defogging method and generator network

The invention discloses an image defogging method of a dark channel attention optimization cyclic generative adversarial network. The method comprises the following steps: establishing a dark channel attention sub-network, establishing generators G1 and G2, and establishing discriminators DX global, DX local, DY global and DY local; calculating a dark channel of a foggy image, inputting the foggy image into the generator G1, inputting the dark channel into a dark channel attention sub-network to obtain an attention map, weighting the middle output of the generator, and finally obtaining a defogged image; adopting a global discriminator to discriminate the defogged image, cutting the defogged image into four small blocks randomly, and adopting a local discriminator to discriminate the defogged image. Haze in the image can be removed, the visual performance of the image is improved, and the problem that the image is blurred under the foggy weather condition is effectively solved. 本发明公开了一种暗通道注意力优化循环生成对抗网络的图像去雾方法