Optic disc detection based on classification for retinal image

A classification method based on Optic Disc (OD) detection is proposed in this paper. It automatically generates a set of optimal features for classifying the OD areas and therefore eliminates the errors due to complicated feature extraction process of previous approaches. Validations on 4 public databases showed the proposed method is superior to existing work on both success rate and speed. The present method achieved an average success rate of 97.8%, (332 images out of 340 testing images). Moreover, OD detection could be completed within 1.04 seconds (need 1.04 ± 0.8) per image.