Preprocessing and Blood Vessel Segmentation of Retinal Images

M.U. Akram, A. Atzaz, M.Y. Javed, and U.A.K. Niazi (Pakistan)


Retinal image preprocessing, wavelet, vessel segmentation.


Automated diagnosis of Diabetic Retinopathy is done by taking retinal images into account. Before detecting the features and abnormalities in the retinal image, preprocessing is required. Retinal image vessel segmentation and their branching pattern are used for automated screening and diagnosis of diabetic retinopathy. We propose a method for color retinal image preprocessing i.e. creating a binary mask to remove the noisy area and background from retinal image. We present a method that uses 2-D Gabor wavelet and sharpening filter to enhance and sharpen the vascular pattern respectively. Our technique extracts the vessels from sharpened retinal image using edge detection algorithm and applies morphological operation for their refinement. The proposed method is tested on publicly available DRIVE database of manually labeled retinal images. The validation of our retinal image preprocessing technique and vessel segmentation is supported by experimental results.

Important Links:

Go Back