Open AccessFusion Image Style Transfer Network
Author(s) -
Shuren Lai,
Fei You,
Hechen Gong,
Yangze Zhao
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1302/3/032002
Subject(s) - computer science , image (mathematics) , stylized fact , artificial intelligence , autoencoder , image fusion , representation (politics) , style (visual arts) , computer vision , fusion , graph , artificial neural network , theoretical computer science , geography , linguistics , philosophy , archaeology , politics , political science , law , economics , macroeconomics
Generating art automatically and quickly for machines is always a difficult task. The existing image style migration algorithms can realize the rapid generation of art images. However, these algorithms can only generate artistic images by migrating existing content maps, and can not modify the content of images artificially. This paper proposes an image style migration approach which can stylize artificially modified fusion image. Particularly, this strategy can correct the modification of the photographic images and enhance the style shifting details. Aiming at eliminating traces of modification, we adopt a autoencoder network to increase fusion image coordination. This network increases detail representation of fusion image. In order to enhance visual effects of style migration, we improve a neural style migration network to generate fusion image stylized graph. We show that this approach make the fused image more coordinated, resulting in a richer detail of the style image.