Eye diseases such as diabetic retinopathy (DR) affect a large number of the population. Retinal fundus photographs are widely used in the diagnosis and treatment of various eye diseases in clinics. It is also one of the main resources for mass screening of diabetic retinopathy. The resulting retinal images must be examined by an expert human grader in a cumbersome and time-consuming diagnosis process. Automated analysis and diagnosis has the potential to reduce the workload and thus increase the cost-effectiveness of such screening initiatives. Nevertheless, there are number of problems that must be solved in order to develop a fully reliable automated retinal images analysis system. Among them, is the need to guarantee that the quality of the retinal images to be graded exceeds a threshold below which the automated analysis procedures may fail [PiOlDa12]. In a DR system, an image is considered poor quality if it is difficult or impossible to make a reliable clinical judgment on the image regarding presence or absence of DR [YuEtAl12]. Performing automated analysis on the image of insufficient quality will produce unreliable results. Images with low quality should be examined by an ophthalmologist and reacquired if necessary [NiAbVa06]. The store and forward teleophthalmology systems involve acquiring images and transmitting them for remote retinopathy detection. This could become problematic when received images do not have enough quality and patient is not accessible

Keywords