Illumination correction of retinal images using superimpose low pass and Gaussian filtering

The digital retinal camera produces a high quality colour digital image and instantly displays it on a large, high resolution monitor but the light reflected by the retinal surface produce a luminosity and contrast variation. In this paper, a new approach for illumination correction by normalizing the background based on a combination between low pass and Gaussian filtering. The experiment had been tested with several selected non-uniform retinal images from Digital Retinal Images for Vessel Extraction (DRIVE) database. Based on Signal Noise Ratio (SNR) result, the proposed method achieved until 14.8020 compared to the original image which obtained 8.2902. In the meantime, the background correction increased up to 78.55%. Finally, the Otsu thresholding was applied to find the binarization images.