Retinal image enhancement using Laplacian pyramidal multi-scaling

Abstract

Early detection of retinal diseases is important to avoid complications and permanent vision loss. In this paper retinal neovascularization and molecular degeneration has been emphasized. Neovascularization is in form of randomly disoriented micro vessels in retina. So image enhancement techniques are excellent way to extract the vessels, find out blood leakages, determine direction of growth and estimate the growth rate with vessel localization. A comparative study has been done on prior retinal image enhancement algorithms. In this project multi-scale image analysis is used as main image enhancement technique with the help of Laplacian Pyramid. The target is achieved by translating an image into several image scales and reconstructing with enhancement tools available in MATLAB image processing toolbox. Results are evaluated with object background contrast ratio, contrast- noise -ratio and 2-D contour plot. The enhanced images appear as a better source for edge detection and vessel extraction compare with the primary image. For this project normal fundus images from publicly available database are chosen.