With the popularization of light-emitting diodes (LEDs) for indoor illumination, visible light communication (VLC) technology using LEDs as optical sources has gained an enormous amount of attention. In most indoor lighting places, lighting systems are composed of a set of LEDs rather than single LED, which is favorable for space-division multiple-access (SDMA) technology. For a SDMA VLC system, one of the major problems is high multi-user interference (MUI), especially in the case that different users have similar geometrical locations. In light of this, a reduced-complexity user-grouping algorithm with mutual spatial correlation as the user-grouping metric is developed. Within a timeslot, only the users with low correlation between each other are served simultaneously, while the users with high correlation are served in different timeslots. Further more, two kinds of optimal linear precoding designs are studied based on the results of user assignment. Groups of simulate results are taken to characterize the performance of both the user-grouping algorithm and the precoding design.