Medical image super-resolution with laplacian dense network
- PDF / 2,663,547 Bytes
- 14 Pages / 439.642 x 666.49 pts Page_size
- 101 Downloads / 332 Views
Medical image super-resolution with laplacian dense network Rui Tang1 · Lihui Chen1 · Rongzhu Zhang1 · Awais Ahmad2 · Marcelo Keese Albertini3 · Xiaomin Yang1 Received: 14 January 2020 / Revised: 8 June 2020 / Accepted: 9 September 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract High resolution medical images are expected for accurate analysis results in medical diagnosis. However, the resolution of these medical images is always restricted by the factors such as medical devices, time constraints. Despite these restrictions, the resolution of these medical images can be enhanced with a well-designed super-resolution(SR) algorithm. As a post-processing manner after medical imaging, the adoption of the SR algorithms has the advantages of low cost and high efficiency compared with upgrading medical devices. In this paper, we propose a network named LDSRN that combines the Laplacian pyramid structure and the dense network to reconstruct clear and convincing medical HR images. Our LDSRN can make full use of the information from different pyramid levels to recover faithful HR images by the dense connection. Specifically, the Laplacian structure decomposes the difficult SR task into several easy SR tasks to obtain the HR images step by step for better reconstruction. Experimental results demonstrate that our LDSRN can obtain better HR medical images than several state-of-the-art SR methods in terms of objective indices and subjective evaluations. Keywords Medical image · Super-resolution · Laplacian pyramid structure · Dense convolutional neural network
1 Introduction High resolution medical images can provide more accurate information compared with the low resolution images to help the doctors conduct accurate analysis. However, the resolution of the medical images is usually limited by medical devices. There are two ways to improve Xiaomin Yang
[email protected] 1
College of Electronics and Information Engineering, Sichuan University, Chengdu, Sichuan, 610064, China
2
Dipartimento di Informatica (DI), Universit`a degli Studi di Milano, Via Celoria 18, Milano MI 20133, Italy
3
Faculty of Computing, Federal University of Uberlandia, Uberlandia, Brazil
Multimedia Tools and Applications
the resolution of the medical images. People can upgrade the medical devices to obtain higher resolution images in hardware methods. However, it is unrealistic in rural areas for the upgrade because it will result in huge cost. Compared with the high cost for expensive medical devices, software methods are actually preferred due to their low cost. Without improving existing conditions, high-resolution (HR) images can be reconstructed directly by using SR methods from its corresponding low-resolution (LR) images. In general, super-resolution algorithms can be divided into three main categories: interpolation-based image SR, reconstruction-based image SR and learning-based image SR. The interpolation-based SR methods obtain the pixel values of the HR image by nonuniform interpolation within
Data Loading...
