Fusion of Entropy-Based Thresholding and Active Contour Model for Detection of Exudate and Optic Disc in Color Fundus Images

Abstract

This paper proposes an efficient and accurate exudate and optic disc (OD) region segmentation methodology. Exudate, an inflammation that occurs in diabetic retinopathy, must be localized for diabetic retinopathy diagnosis. Similarly, the OD region must be inspected for changes in the macular area. Two methods are proposed for locating exudate and the OD region in color fundus images, respectively. The algorithms are then combined to build a single exudate and OD region segmentation algorithm. The methodology uses color normalization to the green channel color space, an intermediate pre-processing step, and a region segmentation step, where active-contour and entropy-based thresholding techniques are applied for segmenting an image to extract exudate and OD. The proposed method is tested on the MESSIDORe, e-ophtha, DIARETDB1, STARE, Pattern Recognition Lab (CS5), and local databases. The segmented images are compared with ground truth images manually generated by a clinician. The segmentation accuracy is found to be 98%. The algorithm successfully delineates the region of interest from the background.