Sie befinden Sich nicht im Netzwerk der Universität Paderborn. Der Zugriff auf elektronische Ressourcen ist gegebenenfalls nur via VPN oder Shibboleth (DFN-AAI) möglich. mehr Informationen...
Journal of physics. Conference series, 2021-04, Vol.1883 (1), p.12057
Ort / Verlag
Bristol: IOP Publishing
Erscheinungsjahr
2021
Link zum Volltext
Quelle
Free E-Journal (出版社公開部分のみ)
Beschreibungen/Notizen
Abstract
Recent state-of-the-art super-resolution methods have achieved impressive performance on ideal datasets. However, these methods always fail in balance between spatial details and high-level perceptual information. Most of them adopt downsampling step to construct Low-Resolution (LR) and High-Resolution (HR) training which may lose local spatial details. To address this issue, we focus on designing a hierarchical attention maps mechanism for recovering both local spatial details and global perceptual information. By using our novel Hierarchical attention module, we can acquire better High-Resolution (HR) predicted images. Finally, we propose a hierarchical multiple scale feature concatenation module aiming at better perception. Extensive experiments on real-world images demonstrate that our method achieved better visual quality both for perception and quantitative estimation.