Feature Extraction for Classification from Images: A Look at the Retina

A method is proposed for the representation of localised features using disjoint sub-images taken from several datasets of retinal images for use within an incremental learning system. A tile-based localised adaptive threshold selection method was taken for vessel segmentation based on separate colour components. Arteriole-venous differentiation was done using the composite of these components and high quality fundal images. Vessel segmentation performance was evaluated on the DRIVE and STARE datasets achieving average specificity of0.9379 and sensitivity of 0.5924.