Constructing Foundations of Super Compressed Sensing and its Applications to Various Tomographic Imaging Modalities

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K06103
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Measurement engineering
• Research Institution: University of Tsukuba
• Principal Investigator: Kudo Hiroyuki 筑波大学, システム情報系, 教授 (60221933)
• Co-Investigator(Renkei-kenkyūsha): OKADA Toshiyuki 筑波大学, 医学医療系, 助教 (90733650)
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: 圧縮センシング / コンピュータトモグラフィー / CT / 画像再構成 / 画像処理 / 電子線トモグラフィー / 逆問題 / スパースモデリング

Outline of Final Research Achievements

Recently, Compressed Sensing (CS) has been successfully applied to image reconstruction in medical x-ray CT and Transmission Electron Microscopy (TEM). In this research project, we propose a new mathematical framework of CS named as Super Compressed Sensing, which significantly improves the performances of CS with respect to image quality. The key of Super CS is to use the nonlinear filter called Nonlocal Mean Filter to evaluate signal sparsity, which leads to preserving complicated intensity changes in images such as image textures and image gradations. We applied the Super CS to image reconstruction in medical CT and TEM. The results demonstrate that the Super CS significantly outperforms the ordinary CS in terms of image quality.

Free Research Field

医用画像工学