Retinal vessel segmentation using the 2-D Morlet wavelet and neural network

This paper proposes a new method for automatic segmentation of the vasculature in retinal images. The method is based on the analysis of feature vectors extracted from a prototype image, to classify pixels as vessel or non-vessel, using a multilayer feed forward neural network. The feature vectors are composed of the pixelspsila intensity and a continuous two-dimensional Morlet wavelet transform of multiple scales. Morlet wavelet has been used because of its ability to tune on specific frequencies, thus allowing noise filtering and vessel enhancement. The classification performance is evaluated by the area under the receiver operating characteristic (ROC )curve, which achieves about 96.68%.