SOBEL OPERATOR AND PCA FOR NEAREST TARGET OF RETINA IMAGES

Abstract

In eye, innermost layer is retina. Various important anatomical structures are available in this. Different eye diseases like diabetic retinopathy, glaucoma, etc are indicated by this. For clinical study, patient screening, and diagnosing ocular diseases, physicians are assisted by vascular intersections and blood vessels extraction in retinal images. Retina image’s nearest template are detected using fuzzy neural network (FNN), Probabilistic neural network (PNN) and Adaptive Neuro Fuzzy Inference System (ANFIS) classifier’s ensemble in recent work. However, various factors like low contrast are having adverse effect on image quality. Accuracy is reduced by this and it is susceptible to errors owing to different human factors. For overcoming these issues, image denoising is introduced first in this work. Clustering algorithm is used for performing this denoising. Fuzzy clustering are used for removing noises present in samples. The, Sobel edge detection operator is used for detecting edges in retina images. To enrich retina image in image enhancement, enhanced linear contrast stretching is used. Mutual Information (MI) optimization is initialized as a coarse localization process using dimension reduction, where local optima are avoided and optimization domain is narrowed down. Then, Enhanced fuzzy neural network (FNN), Improved Probabilistic neural network (PNN) and Adaptive Neuro Fuzzy Inference System (ANFIS) classifier’s ensembling is used for recognizing retina image’s nearest template.

Keywords

• contrast stretching
• diagnosing ocular diseases
• edge detection
• retina image