Reliable in vivo quantification and visualization of myelin spatio-temporal changes in the developing and aging human central nervous system (CNS) would help provide important surrogate markers of white matter integrity. However, high resolution whole brain myelin water fraction (MWF) mapping methods remain challenging and are not well-standardized and are not widely adopted due to ill-posed solutions of the sum-of-exponents problem in the presence of noisy measurements and excessive computation time due to iterative solutions. In this work, we report a rapid method to initialize the solution of the bi-exponential multi-compartment fit which was used along with a host of quality assurance measures to speed up the estimation of the MWF and corresponding compartmental T2 relaxation times using an improved and regularized non-negative least-squares (rNNLS). The sensitivity of the method to signal-to-noise ratio (SNR) and different values of MWF was investigated using a digital phantom. This method was applied to healthy and multiple sclerosis (MS) patients. Our analysis strategy accelerated the accurate mapping of high resolution MWF in the entire human brain in seconds and provided the spatial variability MWF across the corticospinal tract (CST) and callosal pathways.