Retinal Image Analysis Using Contourlet Transform and Multistructure Elements Morphology by Reconstruction

Abstract

Retinal images play a vital role in most of the applications like ocular fundus operations and human recognition. Also, it is used to detect the diabetes in early stages by evaluating all the retinal blood vessels together. The detection of blood vessels from the retinal images is generally a slow process. In this paper, a novel algorithm called Contourlet Transform is proposed to detect the blood vessels efficiently. The proposed Contourlet Transform is the extension of wavelet transform used to enhance the retinal image then the image is utilized for the segmentation part. The existing curvelet transform has disadvantages that is directional specificity of the image is less owing to that the effectiveness is poor. The directionality features of the multistructure elements technique construct it as an effectual tool in edge detection. Therefore, morphology operators by means of multistructure elements are given to the enhanced image in order to locate the retinal image ridges. Later, morphological operators by reconstruction eradicate the ridges not related to the vessel tree as trying to protect the thin vessels that are unaffected. This approach uses multistructure elements in order to improve the performance of morphological operators by reconstruction. An improved Ostu thresholding method is combined with Strongly Connected Component Analysis (SCCA) which indicates the remained ridges pertaining to vessels. The experimental results show the proposed method obtains 96% accuracy in detection of blood vessels and is compared with other existing approaches.