Automatic detection of diabetic retinopathy used in screening systems. In this research a novel splat feature classification method is
proposed for the retinal hemorrhage detection to large irregular hemorrhage detection in fundus images. In proposed method retinal
color image segmented into non-overlapping segments and each segment contain information about pixels and their spatial locations.
A set of features are extracted from each splat to describe its characteristics relative to its surroundings, employs responses from a
variety of filter bank, interacting with neighbor splats and shape and their texture information. An optimal subset of splat features are
selected by a filter approach followed by a wrapper approach. A FNN classifier is proposed to train with splat-based expert
annotations. To improve classification performance with this work FNN classifier is added instead of KNN. A FNN classifier is
proposed to train with splat-based expert annotations. FNN method it is based on pixel classification using a feature vector extracted
from preprocessed retinal images and given as input to a neural network. A variety of lesion detection tasks can therefore be
generalized into exactly the same framework by training classifiers with optimal features learned from available examples projected
onto a sub-feature space which maximizes the inter-class distances while minimizes the intra-class distance.