Blind face images deblurring with enhancement
- PDF / 5,693,766 Bytes
- 21 Pages / 439.642 x 666.49 pts Page_size
- 23 Downloads / 176 Views
Blind face images deblurring with enhancement Qing Qi1,2 · Jichang Guo1 · Chongyi Li3 · Lijun Xiao1 Received: 2 April 2019 / Revised: 7 March 2020 / Accepted: 28 July 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Face images deblurring has achieved advanced development; however, existing methods involve high computational cost problems. Furthermore, the recovered face images by current methods have the problems of over-smooth textures, ringing artifacts, and poor details. We consider the problem of face images deblurring as a semantic generation task. In this paper, we propose a generative adversarial network (GAN), which includes a perceptioninspired blurry removal generator and a discriminator. The proposed generator reconstructs the latent deblurred image by a U-net based network that contains an enhancement module. Face images are highly structured, and thus can be served as a class-specific prior. Considering this, we propose a perceptual loss function to regularize the recovery of face images, which introduces more clear details and reduces the effects of artifacts. The proposed method has a robust capability of generating realistic face images with pleasant visual effects. Extensive experiments on both synthetic and real-world face images demonstrate that the proposed method is comparable with state-of-the-art methods. Keywords Deep learning · Face images deblurring · Enhancement module
Jichang Guo
[email protected] Qing Qi [email protected] Chongyi Li [email protected] Lijun Xiao [email protected] 1
School of Electrical and Information Engineering, Tianjin University, Weijin Road 92, 300072, Tianjin, China
2
School of Physics and Electronic Information Engineering, Qinghai Nationalities University, Bayi Middle Road 3, 810007, Xining, China
3
Department of Computer Science, City University of Hong Kong, Tat Chee Avenue, Kowloon, 999077, Hong Kong, China
Multimedia Tools and Applications
1 Introduction Blind images deblurring task is an ill-posed problem where both sharp latent image and kernels need to be recovered from the single degraded observation. Therefore, additional feature priors are required to constrain image recovery. Single image deblurring task has benefited from hand-crafted priors which are usually developed by natural images and have made advanced progress. Our main focus is on the task of the face images deblurring, the proposed method is potentially applicable in the other types of image. Tackling the face image deblurring problem by directly exploiting classic deconvolution methods which designed generic images, including [51] and the methods of heuristic edge selections are less effective. In other words, these methods mentioned above are modeled on the statistics of natural images and fail to characterize the properties of face images. Furthermore, the strategy of extracting obvious edges to solve deblurring problems is not always effective. In the case that images have large motion blur or highly structured, it is difficult for edges
Data Loading...
