Glaucoma Type Identification from Multiple Fundus Image Modalities Using 2D-FBSE-EWT

Abstract

One of the major causes of blindness is Glaucoma, and accurate classification of different subtypes of glaucoma is important for timely treatment planning. This work presents a multi-fundus image framework using multi-resolution subband image (SBI) features to classify primary glaucoma, secondary glaucoma, and healthy subjects. A two-dimensional Fourier-Bessel series expansion-based empirical wavelet transform has been applied to extract SBIs, and features like fractal dimension, Zernike moments, and several entropy features, are used with machine learning model for classification. Evaluation on a private Indian database showed 94.64% accuracy through feature fusion across different fundus images, full fundus, segment optic disc, and converted polar optic disc images, with the optic disc contributing the most. The good classification results suggest that the framework can be a good tool for assisting ophthalmologist for real time type of glaucoma diagnosis. © 2026 IEEE.