A Novel Retinal Image Color Texture Enhancement Method Based on Multi-Regression Analysis

摘要

Characteristics of retinal tissue shown in retinal images are very important bases for ophthalmologists in clinical diagnosis. Among them, retinal vessel is the most important character. However, the quality of retinal image is usually poor due to the non-perfect imaging environment, making ophthalmologists difficult to read. This paper introduces a retinal vessel enhancement algorithm using multiple regression analysis (MRA) based color transformation scheme to enhance the contrast between retinal vessels and background area in retinal images. This method uses the color texture of another appropriate image as a reference to enforce color transformation on the desired target image. This paper focus on investigating the vessel enhancement effect with different reference images. The proposed method has been validated with STARE (Structure Analysis of the Retina) database. The experimental results show that, using images with lots of green and a little yellow as references can upgrade the average value of contrast to 0.2921 and 3.25 times more than the original one in STARE database. It demonstrates that the proposed method can effectively enhance the contract between retinal vessels and background area in retinal images.