Efficient segmentation and classification of the retinal image enables accurate diagnosis of the Diabetic Retinopathy at the early stages. This prevents the vision loss of the patients. Existing edge detection techniques could not accurately segment the fundus images due to the presence of noise and high-frequency contents. This paper proposes a Hybrid Morphological-based Scanning Window Analysis and Hit-or-Miss Transformation (HMSWA-HMT) and Difference Subspace Sparse representation based Classification (DSSRC) approach to segment and classify the retinal image, for the efficient diagnosis of Diabetic Retinopathy. The morphological operators are applied to minimize the noise or enhance the image. The SWA method provides better localization irrespective of the complex disorder patterns in the image. The HMT detects the significant points such as bifurcations points, ridge and vessel ends that are able to define the vascular skeleton. The Difference Subspace Sparse representation based Classification (DSSRC) method focuses on improving the distinguish ability for the classes rather than the representation capability for the samples. From the experimental analysis, it is observed that the proposed HMSWA-HMT approach achieves better performance in terms of accuracy, sensitivity, true positive rate and specificity and requires minimum time complexity than the existing segmentation and classification techniques.
