An information geometry based scheme for hard exudate detection in fundus Images

Our research focuses on the detection of hard exudates (HEs) in fundus Images for identification of the condition of Diabetic Retinopathy (DR). HEs have been found to be the most specific markers for the presence of retinal oedema, and are also one of the most prevalent lesions during early stages of DR. In order to detect HEs, we have introduced a novel way of using the Distance Learning Metric using a Non-linear Kernels function. We have also introduced a new method to remove the OD in the post-processing stage based on variance calculation. In this the contention in the worst case is resolved by the use of the more robust vertical blood vessel edge detection method. Overall, with the combined implementation of the standard algorithm and the novel algorithm proposed by us, we have been able to achieve classification results of 96.67% (which is better than the results contended in the present literature surveyed by us) with very less computational cost of the algorithm as compared to the computational costs posed by conventional algorithms.