Prior-DualGAN: Rain rendering from coarse to fine

The success of deep neural networks (DNN) in deraining has led to increased research in rain rendering. In this paper, we introduce a novel Prior-DualGAN algorithm to synthesize diverse and realistic rainy/non-rainy image pairs to improve DNN training for deraining. More precisely, the rain streak p...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	Signal processing. Image communication Ročník 129; s. 117170
Hlavní autoři:	Hu, Mingdi, Yang, Jingbing, Yu, Jianxun, Jing, Bingyi
Médium:	Journal Article
Jazyk:	angličtina
Vydáno:	Elsevier B.V 01.11.2024
Témata:	Data-driven deep learning Generative adversarial network Multi-scale convolutional sparse coding Rain rendering Single image deraining Data-driven deep learning 0000 Rain rendering 1111 Generative adversarial network Single image deraining Multi-scale convolutional sparse coding
ISSN:	0923-5965
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Popis
Shrnutí:	The success of deep neural networks (DNN) in deraining has led to increased research in rain rendering. In this paper, we introduce a novel Prior-DualGAN algorithm to synthesize diverse and realistic rainy/non-rainy image pairs to improve DNN training for deraining. More precisely, the rain streak prior is first generated using essential rain streak attributes; then more realistic and diverse rain streak patterns are rendered by the first generator; finally, the second generator naturally fuses the background and generated rain streaks to produce the final rainy images. Our method has two main advantages: (1) the rain streak prior enables the network to incorporate physical prior knowledge, accelerating network convergence; (2) our dual GAN approach gradually improves the naturalness and diversity of synthesized rainy images from rain streak synthesis to rainy image synthesis. We evaluate existing deraining algorithms using our generated rain-augmented datasets Rain100L, Rain14000, and Rain-Vehicle, verifying that training with our generated rain-augmented datasets significantly improves the deraining effect. The source code will be released shortly after article’s acceptance. [Display omitted] The pipeline of the proposed framework. The flow chart on the left illustrates our proposed method, while the illustration of each module is displayed on the right and introduced in detail below. •Construction of rain streak prior as illustrated in Fig.2 ①.•The flowchat of synthesized rain streaks is illustrated in Fig.2 ②.•Further, the flowchat of synthesized rainy images is illustrated in Fig.2 ③.•A new joint discriminator is designed as in Fig.2 ④.
ISSN:	0923-5965
DOI:	10.1016/j.image.2024.117170