A novel hybrid segmentation approach for optic papilla detection in high resolution fundus images of retina

The paper proposes a novel method for segmenting optic papilla (OP) from the high-resolution fundus (HRF) image. For diagnosing eye-related diseases like Glaucoma, Diabetic Retinopathy, fine changes in OP must be examined. To examine the OP, its region should be exactly segmented from the fundus images of the retina. Major problems in accurately segmenting the OP are: 1) features of OP and exudates are similar and 2) the region behind the optic nerve head is difficult to locate. To overcome these problems and acquire precise segmentation a novel hybrid segmentation algorithm is developed using morphological image processing techniques and entropy filtering. A novel region selection algorithm based on Euclidean distance is proposed to remove the regions around the OP. When these regions are removed, the papilla region can be located easily. Then, active contour is applied to segment the OP. Most of the researchers have done OP localization than segmentation. The proposed method automatically locates the OP and segments it from the image. The proposed algorithm is evaluated by computing sensitivity, specificity, and accuracy. These metrics are computed using the proposed segmentation results against the ground truth data. To check the efficiency of the proposed algorithm, it is tested with low-resolution images in DRIONS. The proposed algorithm achieves 99.45% and 99.51% of accuracy for HRF and DRIONS datasets respectively.