Principal Component Analysis in Digital Image Processing for Automated Glaucoma Diagnosis
In digital image processing (DIP), attribute vectors that tend to contain a large number of elements are generated, and some of these elements are irrelevant for image classification. When working with automated classification techniques, some commonly verified attributes may have low relevance for solving a specific problem or even worse the classification, unnecessarily increasing the dimensionality of the problem. Thus, using methods to reduce the dimension of the problem representation space, there may be better interpretation of the data. In this perspective, in this research the Principal Component Analysis (PCA), a widespread dimensionality reduction technique in the literature, was applied to the attribute vector generated by the DIP, aiming to increase the accuracy of the classifier. As a case study, retinal image classification for the diagnosis of glaucoma was used. The results showed a better classification of the images, validating the possibility of applying PCA to optimize the automated glaucoma diagnosis process.