Multicomponent T2 analysis (mcT2) can be highly valuable for probing tissue microstructure. However, it remains challenging due to its ill-conditioned nature, and due to inherent contamination of multi spin-echo signals by stimulated echoes. We present a novel mcT2 algorithm that tackles the high-dimensionality of this problem, using correlations between local and global features of the anatomy in question. The accuracy of this tool is demonstrated on phantoms and in vivo. Our results suggest that the method can accurately identify microscopic compartments, operate at realistic scan times, and be used estimate to estimate myelin content in vivo.