Retinal Image Enhancement Using Curvelet Based Sigmoid Mapping of Histogram Equalization

Abstract

Abstract Ophthalmologists generally use retinal fundus images to identify certain retinal diseases. However, fundus cameras frequently fail to capture high-quality retinal images due to improper camera settings, eye movement, uneven illumination, and pupil dilation that affect the diagnosis’s reliability. To enhance the fundus image’s visual clarity, we propose a combination of denoising and enhancement methods. This paper uses a multi-resolution curvelet transform and adaptive sigmoid mapping of histogram equalization for better image denoising and enhancement. Our hybrid technique enhances the quality of fundus image with improvement in Peak Signal to Noise Ratio (PSNR) of 6.85%, Structural Similarity Index (SSIM) of 0.89%, and Correlation coefficient (CoC) of 0.13%compared to existing methods with Gaussian noise of about 0.01.