Enhancement of fundus imagery

Abstract

Diabetic Retinopathy (DR) is usually diagnosed and graded based on the incidence and severity of micro-aneurysms, hemorrhages, macular/retinal edema andexudates on the fundus images. Similarly, optical Cup to Disc Ratio (CDR) is another feature based on which the progression of glaucoma is evaluated. To enable easy identification of these clinical indications from the fundus images, their contrast has to be improved to appreciable level. Contrast Limited Adaptive Histogram Equalization (CLAHE) is one of the widely accepted contrast enhancement scheme for fundus images. But, in CLAHE, the contrast and quality of the enhanced image heavily relies on the colour model used to represent the contextual image, the channel/component subjected to equalization and the shape of the specified histogram. This article investigate the colour model, adequate channel to be equalized and the histogram specification which offer maximum grey level contrast and image quality in fundus images, when enhanced with CLAHE. It has been observed that RGB model outperforms HSV model, the adequate channel to be equalized is the green and exponential histogram is the apt choice for fundus imagery. The experimental analysis is performed in Matlab®.