Myocardial fibrosis (MF) is a common feature of cardiac disease, characterized by excessive deposition of collagen (i.e., scar tissue) and expansion of the myocardial extracellular volume (ECV). This phenomenon contributes to cardiac dysfunction, promotes further cardiac disease, and has implication in preceding cardiac morbidity and mortality. The extent of myocardial MF can be analyzed globally (across the entire myocardial region) and/or regionally (across the fibrotic area exclusively) using cardiac magnetic resonance (CMR) imaging techniques, such as late gadolinium enhanced imaging or quantitative methods like native T1 and ECV mapping. CMR-based measurements of MF, native T1, and ECV allow for differentiation between various cardiac disease states and are shown to be clinically significant predictors of patient outcomes. However, in order to analyze tissue volumes or classify disease states, clinicians must first perform a manual tracing of the myocardial borders to define an initial region of interest (ROI), while regional MF quantification requires additional manual selection of a reference healthy myocardial tissue region. These manual processes are tedious, user-dependent, and highly prone to operator error, which can significantly confound resultant measures of T1, ECV and quantified MF tissue zones. Thus, alternative, minimally user-dependent techniques for MF, T1 and ECV quantification are appropriate. In this dissertation, several techniques for improving automated quantification of myocardial T1, ECV, and MF regions are presented. The proposed approaches presented in this document incorporate concepts from deep learning and image processing to achieve automated or semi-automated segmentation of the myocardium, MF, T1 and/or ECV in the left ventricle (LV) and left atrium (LA).