Share Email Print

Proceedings Paper

Classification of weak specular reflections in laparoscopic images
Author(s): Bidisha Chakraborty; Jan Marek Marcinczak; Rolf-Rainer Grigat
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Specular reflections are present in the majority of laparoscopic videos. If not considered they will affect all further image analysis and registration algorithms. In most state-of-the-art algorithms, segmentation of specular reflections is done by intensity thresholding. However, the strong reflections are detected but the weak reflections are missed. The proposed method automatically detects the contour boundaries belonging to specular reflections by an SVM classifier. The algorithm improves the detection of small weak reflections by training on contours of specular reflections with a combination of intensity and shape descriptors. Segmentation is done on contours by intensity thresholding and morphological operations. A comparative analysis of the proposed method with the existing methods is presented. The ground truth for the test images is manually labeled for evaluation. The database contains 1012 specular reflections present in 184 images and they are taken from 42 patients. This method improves the sensitivity in detection of weak reflections by 15% as compared to the best known method and 7% for all reflections.

Paper Details

Date Published: 18 March 2014
PDF: 8 pages
Proc. SPIE 9035, Medical Imaging 2014: Computer-Aided Diagnosis, 90353I (18 March 2014); doi: 10.1117/12.2043073
Show Author Affiliations
Bidisha Chakraborty, Technische Univ. Hamburg-Harburg (Germany)
Jan Marek Marcinczak, Technische Univ. Hamburg-Harburg (Germany)
Rolf-Rainer Grigat, Technische Univ. Hamburg-Harburg (Germany)

Published in SPIE Proceedings Vol. 9035:
Medical Imaging 2014: Computer-Aided Diagnosis
Stephen Aylward; Lubomir M. Hadjiiski, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?