Detection of Optical nerve head by using Gabor and Mean filter

The fundus images is manually graded by the specialized clinicians which is very time consuming resource intensive process, so the motivation behind the work is to provide immediate detection and characterization of retinal feature before the specialist inspection. Optical nerve head provide an important anatomical landmark in analysis of retinal disease such as Glaucoma detection, measurement of vessel tortuosity and exudates region and for the screening programs for diabetic retinopathy of prematurity in fundus image of eyes. It is also important landmark to identify the other anatomical parts such as fovea, blood vessels (as it provides the starting point) and other task. The ONH is the entry point for the optical nerves and blood vessel supplying the retina so the center is the point of convergence of main blood vessels .It is present on the nasal side of fundus image .In normal eye it is the brightest, slightly oval disk measuring about 2mm in diameter. The approach used here is based on the detection of vascular tree as a source of reference to locate the ONH. Here after doing the basic preprocessing, in order to detect the vascular tree the resultant is fed to the gabor wavelet. Gabor functions are sinusoidally modulated Gaussian functions that provide optimal localization in both the frequency and space domains. The resultant is then applied to local mean filter of size 30X30. The pixel returning highest intensity value is considered as ONH. This method gave 100% result when applied to DRIVE database and 61.7% when applied to STARE database.