We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Texture Pattern-based Bi-directional Projections for Medical Image Super-resolution.
- Authors
Zhou, Ying; Zheng, Zhichao; Sun, Quansen
- Abstract
The goal of Super-Resolution (SR) is to generate a plausible and visually pleasing High-Resolution (HR) image from a degenerate Low-Resolution (LR) image. High-resolution medical images help improve the accuracy of subsequent operations such as segmentation or remote diagnosis. In this paper, we propose a novel learning-based medical image SR method which directly establishes a bi-directional projection for each texture pattern (texture structures which appear periodically in images) in HR and LR domains. In particular, considering that the feature extraction method should be closely combined with the subsequent steps and the loss of detail is different when different texture patterns degenerate, we propose an auxiliary network to extract features and generate representative texture patterns simultaneously. Then, the textural dictionary is constructed for each texture pattern by introducing texture complexity prior. Hence, the dictionary contains more information when the corresponding texture pattern is complex. Finally, the projections are calculated on these dictionaries with low rank constraint. Extensive experimental results indicate that the proposed method is effectiveness and can deliver higher quality of SR results.
- Subjects
TEXTURE analysis (Image processing); FEATURE extraction; HIGH resolution imaging; DIAGNOSTIC imaging; ENCYCLOPEDIAS &; dictionaries
- Publication
Mobile Networks & Applications, 2023, Vol 28, Issue 5, p1964
- ISSN
1383-469X
- Publication type
Article
- DOI
10.1007/s11036-023-02166-y