Pore-size distribution and permeability can be accurately estimated from nuclear magnetic resonance (NMR) measurements acquired in single-fluid-saturated rocks. However, most rocks penetrated by wells contain multiple fluids and are commonly affected by mud-filtrate invasion, which makes the analysis and interpretation of longitudinal () and transverse () relaxation measurements challenging. It is necessary to replace the hydrocarbon NMR response of the original measurements with an equivalent water response to estimate petrophysical quantities of interest. We have developed an improved NMR fluid substitution method that takes into account partial water and hydrocarbon pore saturation and does not require knowledge of permeability and surface relaxivity. The method consists of two steps: first, the hydrocarbon NMR response is removed from the initially water-hydrocarbon-saturated NMR data. Then, the NMR distribution of the resulting hydrocarbon-depleted system is transformed to that of a completely water-saturated system. Four pore-size-dependent saturation distributions are considered in the study. Tests of verification are performed on Berea sandstone and Indiana limestone samples. Our results show that the method can be reliably applied to Berea sandstone but has limited applicability to the Indiana limestone samples.