Direct manipulation can be used as an intuitive interface for spatial interaction with multiple objects in virtual environments. However, users often have difficulties in precisely manipulating objects in the 3D space; this is mainly because the user is not provided with enough visual information to sufficiently recognize the spatial relationships among objects. We present a new method for improving the preciseness of object manipulation in the 3D space by providing the user with an auxiliary view that is consistently retargeted to closely highlight the spatial relationship between the manipulated object and the object nearest to it. Both the target and the position of the auxiliary camera are computed efficiently using the bisector surface of the two involved objects. The viewing target is set to the center of the closest area between the two objects, and the camera position is optimized under an objective function primarily measuring the visual clarity of the auxiliary view. We demonstrate the usefulness of our method through various experiments based on practical usage scenarios.