Image Inpainting Method Based on Generative Adversary Networks

Abstract

Aiming at the problems of missing and damaged image details, this paper proposes an image inpainting method that is preprocessed and then inpainted. First, through preprocessing that enhances and purifies the image, the clarity, saturation and contrast of the image are improved. Secondly, the image features are dynamically divided through the convolutional neural network, and the image is generated according to the principle of a generative adversarial network, finally, the adversarial strategy is used to promote the expression of the model, to make the repair result more realistic. Compared with other models, the method proposed in this paper has a better quality of repairing effect.