OPTIC DISK DETECTION IN RETINAL IMAGE USING ENERGY MAXIMIZATION ON MATCH FILTER SURFACE

Abstract

THE optic disc (OD) is considered one of the main features of a retinal fundus image, where methods are described for its automatic detection. OD Detection is a key preprocessing component in many algorithms designed for the automatic extraction of retinal anatomical structures and lesions, thus, an associated module of most retinopathy screening systems. The OD often serves as a landmark for other fundus features; such as the quite constant distance between the OD and the macula-center (fovea) which can be used as a priori knowledge to help estimating the location of the macula. The OD was also used as an initial point for retinal vasculature tracking methods large vessels found in the OD vicinity can serve as seeds for vessel tracking methods. Also, the OD-rim (boundary) causes false responses for linear blood vessel filters. A new fast and robust approach to automatically detect the optic disk in the color fundus image is purposed. The purposed method starts by equalizing the contrast through out the image using the adaptive histogram equalization approach. Multiple erosion (cropping) operations are applied in the preprocessing stage in order to shrink image’s ROI. Region of interest (ROI) is selected on the bases of region consisting of high intensity values in the image. So Binarization is applied to extract the region of interest from the image. Dilation is then applied to the ROI to bridge the gapes. Gradient intensity map of the blood vessels is generated by applying the twelve two dimensional, directional matched filters. Further a match filter is generated by averaging the OD vicinity of randomly selected retinal images. Match filter is resized into nine different sizes in order to match the size of the images in the database. Fast Normalized cross-Correlation is applied on to the gradient intensity map and the nine resized match filters. Each correlation surface obtained is convolved with ROI and the refined surface image with the maximum response is selected from these convolved surfaces. In the final step, Extracted peaks in the refined surface image are localized by energy based methodology (block based processing) in order to locate the final optic disk location. Purposed method showed both accuracy and speed. These methods are evaluated using the images of two publicly available databases, the DRIVE database and the DIARETDB0 database. The results of this work are compared with those from other recent methods, leading to the conclusion that our algorithm is comparable with other solutions, while approximating the average accuracy of a human observer without a significant degradation of sensitivity and specificity, with significant improvement in processing time.