Title: Retinal vessel segmentation using an improved multi-scale line detection

Authors: Xiangjun Gao

Addresses: School of Computer and Information Technology, Shangqiu Normal University, Shangqiu, China

Abstract: This paper proposes an effective retinal vessel segmentation method based on a combination of line detectors at different scales. To decrease the influence of optic disc and enhance retinal vessel (especially small vessel) response for each retinal image, the proposed method first applies a top-hat transform and a line filter to the background homogenised retinal image. Then an improved multi-scale detection is used to produce a vascular response image for each retinal image. Since the contribution of different lengths of line detectors to vessel response is not equal, the improved multi-scale detection combines all line responses at varying scales by their contribution to vessel response. The performance of the proposed method is evaluated on two publicly available DRIVE, STARE databases. Experimental results demonstrate that our method achieves high accuracy and approximates the average accuracy of a human observer without a significant degradation of sensitivity and specificity.

Keywords: retinal images; vessel segmentation; line detectors; top-hat transform; image segmentation; vascular response images; multi-scale detection.

DOI: 10.1504/IJBET.2013.058445

International Journal of Biomedical Engineering and Technology, 2013 Vol.13 No.3, pp.240 - 256

Received: 16 Jul 2013
Accepted: 03 Nov 2013

Published online: 27 Sep 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article