Denoising methods for retinal fundus images

• Main Authors: Hani, A.F.M., Soomro, T.A., Faye, I., Kamel, N., Yahya, N.
• Format: Conference or Workshop Item
• Published: IEEE Computer Society 2014
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-84906335959&doi=10.1109%2fICIAS.2014.6869534&partnerID=40&md5=2e22ac1df3f69588f9cbe96b81441eef; http://eprints.utp.edu.my/32150/
• Institution: Universiti Teknologi Petronas

Diagnosing retinal diseases of the eye requires analysing tiny retinal vessels. Digital colour fundus images are plagued by the problem of low and varied contrast. Further, with noise being present in the images, retinal vasculature is difficult to be analysed. This paper discusses various denoising methods to improve the SNR of the retinal fundus images before further image enhancement. By selecting a suitable method for denoising, the image's details are not lost as well as the contrast is maintained. Based on the performance of several techniques for denoising fundus images, it was found that the Time Domain Constraint Estimator (TDCE) showed a greater performance in the PSNR improvement of retinal fundus images. By reducing noise using TDCE, the performance of non-invasive methods for enhancing fundus images can be significantly improved without any loss of the details of the images. © 2014 IEEE.