Optimal Anisotropic Guided Filtering in retinal fundus imaging: A dual approach to enhancement and segmentation

Retinal vascular tree segmentation and enhancement has significant medical imaging benefits because, unlike any other human organ, the retina allows non-invasive observation of blood microcirculation, making it ideal for the detection of systemic diseases. Many traditional methods of segmentation and enhancement encounter issues with visual distortion, ghost artifacts, spatially inconsistent structures, and edge information preservation as a result of the diffusion of spatial intensities at the edges. This article introduces an Optimal Anisotropic Guided Filtering (OAGF) framework tailored for retinal fundus imaging, addressing both enhancement and segmentation needs in a unified approach. The proposed methodology consists of three stages, in the first stage, we perform the illumination correction and then convert the source RGB image to YCbCr format. The luminance (Y) component is further processed through OAGF. In the second stage, optimized top-hat transform and homomorphic filtering has been performed to get segmented image. In the third stage, the enhanced image is produced by converting YCbCr to RGB format. To validate the effectiveness of the suggested approach, extensive experiments with the open-source DRIVE and STARE datasets were performed. Quantitative and qualitative assessments prove that the OAGF-enhancement and segmentation methodology surpasses current algorithms with better values in Dice Coefficient (0.860, 0.854), Precision (0.845, 0.834), and F1 Score (0.827, 0.817) on both databases.