Joint image-to-image translation with denoising using enhanced generative adversarial networks

Impressive progress has been made recently in image-to-image translation using generative adversarial networks (GANs). However, existing methods often fail in translating source images with noise to target domain. To address this problem, we joint image-to-image translation with image denoising and...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Signal processing. Image communication Ročník 91; s. 116072
Hlavní autori: Yan, Lan, Zheng, Wenbo, Wang, Fei-Yue, Gou, Chao
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Elsevier B.V 01.02.2021
Predmet:
Generative adversarial networks
Image denoising
Image enhancement
Image-to-image translation
Image-to-image translation
Generative adversarial networks
Image enhancement
Image denoising
ISSN:0923-5965, 1879-2677
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:Impressive progress has been made recently in image-to-image translation using generative adversarial networks (GANs). However, existing methods often fail in translating source images with noise to target domain. To address this problem, we joint image-to-image translation with image denoising and propose an enhanced generative adversarial network (EGAN). In particular, built upon pix2pix, we introduce residual blocks in the generator network to capture deeper multi-level information between source and target image distribution. Moreover, a perceptual loss is proposed to enhance the performance of image-to-image translation. As demonstrated through extensive experiments, our proposed EGAN can alleviate effects of noise in source images, and outperform other state-of-the-art methods significantly. Furthermore, we experimentally indicate that the proposed EGAN is also effective when applied to image denoising. •We study a novel image-to-image translation with noise problem, this is, translating a noisy image in one image domain to a noise-free image in another image domain.•A trainable end-to-end pipeline and a robust GAN-based model named EGAN are proposed to address this problem.•We experimentally demonstrate the effectiveness of the proposed pipeline and the superiority of EGAN compared to a range of GAN-based image-to-image translation methods.
ISSN:0923-5965
1879-2677
DOI:10.1016/j.image.2020.116072