Retinal Blood Vessel Extraction using Wavelet Transform and Morphological Operations

Retinal image plays a vital role in medical image processing for the analysis of micro vascular diseases like diabetic retinopathy and glaucoma. This research work proposes a vessel segmentation algorithm that can be used in computer aided retinal image analysis. The proposed method comprises of three main stages. In the first stage, preprocessing was done that comprises of green channel component extraction, filtering by decision based median filter and enhancement by adaptive histogram equalization. Then the morphological operations were applied for the elimination of optical disc and background before the extraction of blood vessels. Finally the binarization technique and discrete wavelet transform was applied for extraction of blood vessels and extracted blood vessel bundles were traced on the input fundus image. The quantitative analysis of the result was done by performance metrics like dice coefficient, rand index, variation of information and global consistency error. The algorithms were developed in MATLAB 2010a and tested on real time fundus image from STRIVE database comprising of healthy, diabetic retinopathy and glaucoma cases.