An efficient algorithm for extraction of anatomical structures in retinal images

Abstract

This paper presents efficient methods for automatic detection and extraction of blood vessels and optic disc (OD) both of which are two prominent anatomical structures in ocular fundus images. The blood vessel extraction algorithm is composed of four steps, i.e., matched filtering, local entropy-based thresholding, length filtering, and vascular intersection detection. The OD identification involves the efficient active contour (snake) method. Specifically, we use intensity variation and local contrast to initialize the snake model. Simulation results on a set of retinal images verify the effectiveness of the proposed methods.