Performance Evaluation of Various Filtering Techniques for Retinal Fundus Images

De-noising of the Retinal Fundus images is a necessary pre-processing step in diabetic retinopathy, glaucoma, cardiovascular diseases treatment. This step guarantees ample quality for the Computer-Aided Diagnosing (CAD) systems. For the detection of haemorrhages, well known by the type lesions in diabetic retinopathy early identification is very useful. This paper presents an evaluation strategy for various de-noising filters, which have been affected by salt & pepper noise. In general, the Retinal Fundus images are mostly affected by salt&pepper noise and speckle noises. We have considered the salt & pepper noise for the evaluation approach, 50% of the salt & pepper noise have been added into the Retinal Fundus images. Then the noise has been removed by using various filters like Mean, Median, Weiner, Gaussian, and Adaptive Median Filters. The performance of the above filters is compared on the basis of the parameters like MSE, PSNR, and SNR. This paper concludes that among the filters evaluated, Gaussian filter is the best for removing noises in retinal fundus images and thereby it contributes well for diabetic retinopathy, glaucoma, cardiovascular disease to guarantee ample quality for the Computer-Aided Diagnosing (CAD) systems.