Diabetic Retinopathy is one of the main cause for the damage of retina due to complication of diabetes which can eventually lead to blindness. Features like retinal blood vessels and exudates indicate the presence of diabetes in an individual. In this paper, a method to compare two transforms namely Wavelet and Curvelet is proposed. By feature extraction, area of blood vessels and exudates is obtained. Transforms are applied on it to decide which is a better transform for detection of diabetes. This technique is then tested on FUNDUS retinal images. According to our technique, Curvelet is a better transform to detect diabetes and provides an efficiency of 75%.