Supervised Learning Approach for Performance Evaluation of Digital Fundus Image Quality

The fundus image quality is an important parameter in medical applications. Low quality of fundus images can affect to perform a correct diagnosis. In earliest method, to estimate the fundus image quality in local image analysis technique. Drawback of this method it is only applicable for small dataset and affected some factors in retina features. To prevent those limations we proposed a new method to determine the quality of retinal image whether it is good or bad. Initially preprocessing stage has been employed and calculates the mask.To estimate the focus measures are extracted from each image, then all focus measure values are plotted into box plot because we can predict the rules and assigned to the classifier and classify it. Finally, to obtain the classifier performance is achieved by receiver operating characteristic (ROC) curve.