| Paper Abstract and Keywords |
| Presentation |
2014-11-26 17:05
Visual-Word Feature Transformation Architecture for Computer-Aided Diagnosis using Colorectal Endoscopic Images with NBI Magnification Koki Sugi, Tetsushi Koide, Anh-Tuan Hoang, Takumi Okamoto, Tatsuya Shimizu, Toru Tamaki, Bisser Raytchev, Kazufumi Kaneda, Yoko Kominami, Shigeto Yoshida, Shinji Tanaka (Hiroshima Univ.) VLD2014-77 DC2014-31 |
| Abstract |
(in Japanese) |
(See Japanese page) |
| (in English) |
With the increase of colorectal cancer patients in recent years, the computer-aided diagnosis (CAD) system which supports doctor's diagnosis effectively in the diagnosis of colon cancer is required. Our research describe the support system for colorectal endoscopic images with narrow band imaging (NBI) magnification which identify the pathology type from local features in NBI endoscopic image. In order to meet the real time processing in which a full high definition (Full HD) endoscopic image must be identified within one second, the feature transformation which change feature quantity to VW histogram is implemented on hardware. For implementation on FPGA which has the resource limitations, we propose the architecture that uses Manhattan distance in class searching, and enables high speed processing by dynamic computation of normalization. As a result, it allows the feature transformation to complete its work in about 334 msec on the full HD image. More over, the architecture can improve the processing time to about 43 msec by pipelining. |
| Keyword |
(in Japanese) |
(See Japanese page) |
| (in English) |
Colorectal Endoscopic Images / Full HD / Computer-Aided Diagnosis(CAD) / NBI(Narrow Band Imaging) / Feature Transformation / Visual-Word(VW) / Manhattan distance / FPGA Hardware Implementation |
| Reference Info. |
IEICE Tech. Rep., vol. 114, no. 328, VLD2014-77, pp. 33-38, Nov. 2014. |
| Paper # |
VLD2014-77 |
| Date of Issue |
2014-11-19 (VLD, DC) |
| ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
| Download PDF |
VLD2014-77 DC2014-31 |