No Detail Found!

On the Performance Analysis of Blood Vessel Extraction from Fundus Images using Digital Image Processing.

Author(s)
R. A. KHAN Department of Computer Systems Engineering, University of Engineering & Technology Peshawar, Pakistan
N. MINALLAH Department of Computer Systems Engineering, University of Engineering & Technology Peshawar, Pakistan
Abstract
The pattern, bifurcations, radius of vessels present in fundus retinal images are important pointers of various retinal diseases such that diabetic retinopathy, macular degeneration, arteriosclerosis, hypertensive retinopathy and glaucoma. In this paper we propose an accurate and simple scheme to segment almost all of the vessels present in fundus image thereby using only green channel of RGB fundus image. Contrast adjustment through sigmoid function which plays a key role in our scheme, followed by averaging filtering for background exclusion, hysteresis thresholding and morphological processing for post processing to make segmented vessels more clear and visible. Increasing the segmentation accuracy and keeping computational complexity low, is a huge tradeoff. So our main objective is to segment the large as well as thin vessels with low computational overhead and most accurate manner as compared to other existing methods with large and complex algorithms. The proposed method is tested on all of the fundus images in DRIVE and STAREdatabases. The results drawn are compared and evaluated with the segmentation results of other authors, and our scheme proved to be better. Performance evaluation metrics were taken in to account with major focus on segmentation accuracy.
Publication Details
Page(s) 797-802
DOI DOI not available
Published Journal: Sindh University Research Journal, Volume: 48, Issue: 4, Year: 2016

Request the full text from PASTIC

Keywords
Diabetic Retinopathy, fundus, support vector machine (SVM), Gaussian mixture model (GMM), Contrast limited adaptive histogram equalization (CLAHE) Markov's random field (MRF)
0

Citations

0

Downloads

59

Views