Retinal Image Segmentation Using Gabor Transform with Preprocessing and Hysteresis Thresholding

Scientific study revealed that eyes are the best indicators of many diseases like glaucoma, diabetic retinopathy, hypertension, and stroke. By examining the segmented retinal blood vessel network, ophthalmologist can get the information regarding the abnormality. The objective of this research is to provide reliable segmented retinal blood vessel to assist the ophthalmologists to figure out the abnormality precisely. In this work, enhanced Gabor filter in multiple orientation is employed for the retinal blood vessel extraction. The suggested method combines sharpening operation and median filtering with Gabor transform to get enhanced Gabor transformed images. Finally, to get the segmented output hysteresis thresholding is applied on the enhanced Gabor transformed images. As hysteresis thresholding takes into consideration the connectedness between neighboring pixels, it performs better in segmenting the vessels. The suggested integrated approach has improved the accuracy and specificity. Experiment on 20 retinal images of DRIVE database indicated 95.06% accuracy and 98.83% specificity.