NLOQO Home · Issue Contents · Forthcoming Papers

Enhanced Retinal Image Based Segmentation and Deep Learning
Nedaa Monther Salman, Hazim G. Daway and Jamela. A. Jouda

Retinal images are susceptible to various quality problems including lighting, noise, etc., which reduces anomalies’ visibility. Therefore, improving retinal fundus imaging is crucial for increasing the accuracy of eye disease prediction. In this study, we suggest a new algorithm based on employing K-means and CLAHE technique to segment and enhance retinal images The lighting component is separated from the color components using space YC_bC_r and it is improved and the noise is removed from it using de-noising deep neural network, and then the inverse transform is used to get the improved image, the 40 images from driver data using in this study. Different strategies were contrasted with this one. According to the results, the suggested strategy produced the best average values for the entropy (7.4304), contrast enhancement measurement (0.5861), and naturalness image quality evaluator (5.1727).

Keywords: Enhancement image, RGB image retina, quality assessment, Deep learning K-means, De-
nosing CNN, YC_bC_r