BSEMD Based Automatic Microaneurysm Localization from Fundus Image

Diabetic retinopathy [1] is one of the major implications of diabetes which not only changes the color of blood vessels but also weakened or blurred or wrenched the power of vision and these results blindness. Microaneurysm localization is very essential at early stage so that it could not turn in Diabetic Retinopathy. The fundus images are extensively used for the diagnosis as well as for the treatment of all possible types of eye diseases. Proper localization of microaneurysm is a main solution to prevent from the diabetic retinopathy. However, this screening takes a lot of time, lot of technical persons as well as the proper involvement of doctors. Hence, the need of automated test is widely used now-a-days for the fast and accurate detection. Here this work is done using Bi-dimensional Statistical Empirical Mode Decomposition [2], [3], [12], which not only is a filtering procedure but also it decomposes an image into 2D Intrinsic Mode Functions. The images are first taken in green channel from where the BSEMD is applied on DRIVE database [4]. BSEMD makes the images more smooth and noise free and increases the clarity. This paper applies a decomposition based algorithm based on digital signal processing and image processing to precise diagnose the diabetic retinopathy from images of retina. The result shows its upliftment in terms of proper localization.