Využití metod zpracování obrazu pro úpravy snímků sítnic

Abstract

This bachelor thesis deals with the use of digital image processing methods to improve the quality of retinal images, which are crucial in the diagnosis of retinopathy of prematurity. The introductory part of the thesis describes the different stages of retinopathy of prematurity and the reasons why it is important to take good quality retinal records. In the theoretical part, different image processing techniques such as histogram equalization, CLAHE, color channel extraction, linear and nonlinear filtering, morphological operations and also modern image enhancement methods based on deep neural networks are analyzed. The practical part deals with the application of these methods to real retinal recordings, including the composition of the modified channels and the extraction of vascular structures. The results are evaluated both visually and using quantitative metrics such as SSIM or standard deviation, with a focus on assessing the improvement in image readability and the medical benefits of these enhancements.