An Effective Algorithm for Hemorrhages Detection in Retinal Images

Periodic screening for Diabetic retinopathy (DR) reduces the progress of disease and also prevent loss of visual capability of a person at early stage. This can be achieved using automatic detection of abnormality in DR. To achieve this two methods are proposed for hemorrhages detection in retinal images. Firstly, detect the optic disc for easier segmentation of hemorrhages. After removal of optic disc H-maxima Transform and thresholding approaches are proposed for Hemorrhages(HAs) detection. The proposed algorithms are applied on the DIARETDB1 retinal image database. Receiving operating curve (ROC) and weighted error rate (WER) are used to evaluate the results of proposed algorithm. The quantitative analysis of these parameters shows better performance for thresholding approach to detect HAs.