In the automatic diagnosis of numerous ophthalmic and cardiovascular illnesses, segmentation of retinal blood vessels plays a significant role. Medical analysis and the detection of associated disorders depend on the ability to segment thin and thick retinal vessels. The fundus image of the eye exhibits variable vascular thickness, the latter of which is not segregated by the majority of the previously suggested algorithms. This technique aims to distinguish thick and thin vessels in order to produce two images individually, which can then be combined to provide the final segmentation result. The process includes a number of steps. In addition to applying filters for input retinal image smoothing, primary processing entails many stages. The first one uses an updated top-hat, homomorphic filter to segment large vessels, while the other one presents an optimal top-hat, homomorphic filtering, gray-level conversion, matching filter, and segmentation using multilevel algorithm to segment narrow vessels. Morphological image operations are lastly performed. With this suggestion, a minimal number of false positives are obtained. This technique effectively divides the vessels and raises the performance standards. Implementation part was done in MATLAB R2022b. DRIVE datasets were used to evaluate the suggested method.