Composite Filtering and Morphology for Optic Disc Segmentation from Retinal Images

Optic Disc is one of the important attributes which signifies the status of Different eye related diseases like Diabetic Retinopathy, Glaucoma etc. For this purpose, accurate segmentation of Optic Disc from retinal images is essential. Towards such prospect, this paper proposes a new Optic Disc segmentation mechanism based on Composite Filtering and Mathematical Morphology. The complete methodology is accomplished under two-fold; they are (1) Optic Disc Pixel (ODP) identification and (2) Optic Disc Segmentation. In the first fold, ODP is located as an average of candidate pixel locations which are obtained through three different filters namely Median, Gaussian and Variance filters. Next, the final Optic Disc is segmented after the removal of blood vessels with the help of Linear and Rotational Structuring element. For experimental validation, two standard datasets namely DRIVE and MESSIDOR are used, and the performance is measured through Accuracy, sensitivity, specificity and overlap score. The average sensitivity, specificity and overlap score of proposed method is observed as 0.8865, 0.8722, and 0.8472 respectively. Further, the average improvement in the accuracy gained by proposed method is observed as 5.31%.