Contourlet transform-based sharpening enhancement of retinal images and vessel extraction application

Abstract

Low-contrast retinal images have to be enhanced for good visual perception to aid in retinal vessel analysis. Classical sharpening enhancement techniques such as unsharp masking (USM) improve the contrast and bring out the information along with noise. This article uses a shift-invariant anisotropic Contourlet transform (CT) to decompose the retinal image into subbands. A new nonlinear method is applied over the subbands to modify the CT coefficients, followed by inverse CT. The proposed method is compared with a nonlinear USM (NLUSM) technique and wavelet transform-based method. The objective performance is measured in terms of enhancement measure. We observed that the proposed methodology provides better result. We demonstrate that this sharpening algorithm can be used as a preprocessing step to (i) adaptive histogram equalization and (ii) retinal vessel extraction. Pratt's figure of merit was used to analyze the vessel extracted from the retinal images with their ground truth that were obtained from STARE and DRIVE databases.