Single Image Super-Resolution for MRI Using a Coarse-to-Fine Network
| Autor: | Huabei Shi, Hongen Liao, Jia Liu, Fang Chen |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
Mean squared error
business.industry Computer science Resolution (electron density) ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Pattern recognition Function (mathematics) Image (mathematics) Coarse to fine Range (mathematics) Computer Science::Computer Vision and Pattern Recognition Segmentation Artificial intelligence Single image business |
| Zdroj: | IFMBE Proceedings ISBN: 9789811075537 |
| DOI: | 10.1007/978-981-10-7554-4_42 |
| Popis: | Single Image Super-Resolution (SISR) which aims to recover a high resolution (HR) image from a low-resolution (LR) image has a wide range of medical applications. In this paper, we present a novel Super-Resolution Coarse-to-Fine Network (SRCFN) that recovers the finer texture details strongly and enables precise high-frequency detail to address this challenging task. First, we apply some residuals units to achieve a coarse Super-Resolution result. Second, we add a fine module using the idea of segmentation networks to combine more high-frequency detail into the coarse results for final Super-Resolution results. In addition, we use a combined loss function of Mean square error loss and SSIM loss. Our proposed method applied to medical MRI outperforms previous methods of accuracy (PSNR and SSIM) and visual improvements. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|