Automated Pre-Processing of Retinal Images

Digital retinal images are getting more interest while developing automated systems. These images are of poor quality due to patient’s movement, poor focus, bad positioning, reflections, inadequate illumination, non-sufficient contrast, presence of noise, etc. Pre-processing is necessary to improve the quality of retinal images. In this work, image processing algorithms are implemented for pre-processing steps. The algorithms are developed for boundary detection, colour normalization, colour space conversion, de-noising and contrast enhancement. Normally, contrast enhancement is achieved using histogram equalization method. However, this method over-enhances the central part and the optic disc region of retinal image. In this work, a modified histogram equalization method is adopted for contrast enhancement. The images acquired are used as input images and pre-processing steps are applied sequentially to them. The retinal images often need visual enhancement prior to digital analysis for pathological risk or damage detection. Pre-processing is an essential step for reliable extraction of features and detecting abnormalities.