Applying a Set of Gabor Filter to 2D-Retinal Fundus Image to Detect the Optic Nerve Head (ONH)

Author(s):

Rached Belgacem*, Hédi Trabelsi, Ines Malek and Imed Jabri

Background: The detection of edge of the cup (excavation), disc, and the optic nerves head which converges into the center of the excavation in the papilla of the retina is a crucial procedure in computer vision object detection processing and analysis. Purpose: We propose a method and approach for cup and disc detection based on 2-D Gabor filter with the addition of some constraints. This paper approves that edge magnitude and edge phase are useful for extracting the disc and the cup and to overcome the lack contrast between the cup-disc in order to discriminate one from the other and to extract the different optical vessels along several directions theta. These methods are based on space image and phase image after searching the favorable parameters (Scale, sigma, a & b, r ...) to properly adjust the results of the application of kernels Gabor on the fundus image of the retina to calculate the magnitude image and the phase image which each bring additional information on the excavation of the of the optic nerve head ONH to help to early glaucoma disease screening. Results: The proposed method was applied on several retinal images issued from a set of images of Tunisian glaucomatous database. A total of 60 images were used. For an input image, several features were determined, such as edge of the disc, edge of the cup, boundaries of the papilla and calculate magnitude image and phase’s image. Fitting sigma, scale, spatial frequency F0 and (a, b) parameters were adjusted to aggregate the Gaussian kernel function modulated by a sinusoidal plane wave, ωr (x, y) and C (x, y) a complex sinusoid, respectively. The Signal-to-Noise Ratio SNR and the mean square error MSE were used to identify the optimum fitting parameters instigated in the Gabor function and convolved with the retinal image to detect the magnitude and phase image over different orientations for the segmentation of cup and the disc and the optic nerve head ONH. Conclusions: 2D Gabor filters are analyzed and the apply a set of Gabor filter to 2Dfundus image of retina is efficacious to detect the optic nerve head ONH and the Gabor filter banks for texture classification of the papilla and pattern analysis of the cup and the Disc. The results given in this paper can be useful by an ophthalmologist to diagnosis a glaucomatous case. Segmentation of the cup and the disc through the bank Gabor filter proved functional for glaucomatous diagnosis and screening the pathology.