Image Enhancement Methods For Fundus Retina Images

Early detection of a lesion in the retinal image is becoming very important in order to reduce human misclassification for Diabetic Retinopathy (DR) detection and also helping the ophthalmologists to reduce their burden. In DR, the earliest lesion presents in the retinal image is the microaneurysms (MAs). In this paper, the main goal is to propose the best pre-processing combination method to enhance the retinal image. In this work, seven pre-processing steps were combined and presented in pre-processing stage. Two dataset namely Messidor and E-optha were used as the input image. The image within the dataset were divided into two classes namely, dark image and bright image. Based on the result, the best value of power law function is set to 0.7 for the dark image and 0.9 for the bright image followed by the log transformation as the final method in the pre-processing stage.