Image inpainting based on improved deep convolutional generative adversarial networks

Huaibo Sun; Xiangxiang Han; Yan Zhang; Shan Gao; Xintong Ge; Zeyu Sun

doi:10.1117/12.2656824

20 October 2022 Image inpainting based on improved deep convolutional generative adversarial networks

Huaibo Sun, Xiangxiang Han, Yan Zhang, Shan Gao, Xintong Ge, Zeyu Sun

Proceedings Volume 12451, 5th International Conference on Computer Information Science and Application Technology (CISAT 2022); 124515M (2022) https://doi.org/10.1117/12.2656824
Event: 5th International Conference on Computer Information Science and Application Technology (CISAT 2022), 2022, Chongqing, China

Abstract

Image inpainting based on the deep convolutional generative adversarial networks (DCGAN) is a new technology. However, its inpainting effect is unsatisfactory due to its shortcomings in gradient disappearance and unstable training state. Given this, we propose an image inpainting technology based on an improved DCGAN (IDCGAN), which can ensure the discrimination accuracy and avoid the problems of unstable training and gradient disappearance by optimizing the usage strategy of the activation function (LeakyReLU(), Tanh() and Sigmoid()). In addition, the training effect is improved by redesigning the network, increasing the depth of the network, and using the convolutional kernels flexibly. The experimental results demonstrate that, under the same times of iterations (200 times), the inpainting results of the proposed technology in this paper are up to 40% and 11.48% higher than the reference algorithms in terms of PSNR and SSIM, respectively, and do not affect the impressions of the repaired images at all. Thus, better inpainting performance is achieved.

Citation Download Citation

Huaibo Sun, Xiangxiang Han, Yan Zhang, Shan Gao, Xintong Ge, and Zeyu Sun "Image inpainting based on improved deep convolutional generative adversarial networks", Proc. SPIE 12451, 5th International Conference on Computer Information Science and Application Technology (CISAT 2022), 124515M (20 October 2022); https://doi.org/10.1117/12.2656824

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available