Two-Step Image Hallucination and Its Application to 3D Medical Image Super-resolution

DOI

10.2991/aiie-15.2015.91How to use a DOI?

Keywords

image restoration; super-resolution; medical volumetric image

Abstract

In medical diagnosis, high resolution (HR) images are indispensable for giving more correct decision. However, in order to obtain high resolution medical images, it is necessary to impose long-time, hence it leads to heavy burden to the patient. Therefore Super Resolution technique, which can generate high resolution images from low resolution images using machine learning techniques, attracts hot attention recently. Therein, face hallucination is one of widely used super-resolution methods in image restoration field. However, the conventional face hallucination generally cannot recover high frequency information. Therefore, this paper integrates a further learning step into the conventional method, and proposes a 2-step image hallucination, which is prospected to recover most high frequency information lost in the available low-resolution input. Furthermore, we apply the proposed strategy to generate the high-resolution Z-direction data using self-similarity among different direction for 3D medical MR images. Experimental results show that the proposed strategy can reconstruct promising HR coronal or sagittal plane by using available LR and HR data pairs in axial plane.

Copyright

© 2015, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Y. Kondo
AU  - X.H. Han
AU  - X. Wei
AU  - Y.W. Chen
PY  - 2015/07
DA  - 2015/07
TI  - Two-Step Image Hallucination and Its Application to 3D Medical Image Super-resolution
BT  - Proceedings of the 2015 International Conference on Artificial Intelligence and Industrial Engineering
PB  - Atlantis Press
SP  - 326
EP  - 329
SN  - 1951-6851
UR  - https://doi.org/10.2991/aiie-15.2015.91
DO  - 10.2991/aiie-15.2015.91
ID  - Kondo2015/07
ER  -