Retinal vessel trees play an essential role in clinical decision and quantitative analysis of ophthalmopathy. The changes in the morphological properties of retinal vessels provide accessible information used for diagnosing, screening, and monitoring many eye-related and systemic diseases. Due to the complexity of vessel structures, the manual annotation of vessel network by human experts is a labour-intensive, time-consuming, and error-prone process. Therefore, the automated and accurate delineation of retinal vessels is highly desirable. However, this is a difficult task because of multiple challenges such as the inadequate contrast between vessels and background, variability of vessel widths, uneven background illumination, crossing and bifurcation regions, tiny vessels, and the presence of noise and pathological regions. In this thesis, we aim at introducing novel theoretical concepts that advance the delineation technology of retinal blood vessels to address the shortcomings in existing methods. [...]