Retinal image enhancement using low-pass filtering and α-rooting

Abstract The blurring of a retinal image can be caused by refractive medium turbidity or imperfect imaging conditions, which can decrease the diagnostic reliability. High-quality retinal images are essential for both clinical and computer-aided diagnoses. In this study, a novel method of enhancing retinal images is proposed. It includes four steps, namely padding, contrast improvement, grayscale adjustment, and refinement. Background padding is first operated to prevent an over enhancement of the retinal boundary. Thereafter, we improve the contrast by removing the low-frequency in the root domain. After the contrast improvement, the image color often changes considerably. Thus, a grayscale adjustment for each channel is conducted to recover the original color. Finally, a refinement is used to further enhance the contrast. Four state-of-the-art methods are selected for a comparative study, and the proposed method outperforms the other methods in both a visual assessment and a quantitative analysis. The results indicate that the proposed method can effectively improve the clarity of a retinal image without introducing a considerable color difference.