Low salinity water flooding is a common technique for enhancing oil recovery; however, the mechanism behind the low-salinity effect, positive or negative, is still not fully understood. In the proposed mechanisms, osmosis and emulsification are considered as two potential reasons for explaining the oil remobilization, but the specific contributions on the remobilization are not well studied at pore-scale. In this article, we performed a series of microfluidic experiments to investigate the movement of constrained oil between invading low-salinity brine and residual high-salinity brine. We find that various salinity contrasts over oil films cause different water fluxes through the oil and swelling areas of the trapped brine, resulting in the relocation of oil phases within the pore spaces. A higher salinity contrast (1.7-170 g/L salt concentrations) provides a faster water penetration in oil phases. In the presence of an oil-soluble surfactant, spontaneous emulsification occurs at the interface of low-salinity brine/oil, which enhances almost 100 times the water flux in two oil phases (n-heptane and n-dodecane). We directly observe pore-scale spontaneous emulsification at the low-salinity brine/oil interface but not at the high-salinity brine/oil interface. Furthermore, two scenarios for explaining water transport through the oil phase are proposed: water diffusion due to chemical potential gradient and water transport via reverse micelle or microemulsions movement.