Object occlusion is a major handicap for efficient interaction with 3D virtual environments. The well-known World in Miniature (WIM) metaphor partially solves this problem by providing an additional dynamic view point through a hand-held miniature copy of the scene. However, letting the miniature show a replica of the whole scene makes WIM metaphor suitable for only relatively simple scenes due to occlusion and level of scale issues. In this paper, the authors propose several algorithms to extend the idea behind the WIM to arbitrarily complex scenes. The main idea is to automatically decompose indoor scenes into a collection of cells that define potential extents of the miniature replica. This cell decomposition works well for general indoor scenes and allows for simple and efficient algorithms for preserving the visibility of potential targets inside the cell. The authors also discuss how to support interaction at multiple levels of scale by allowing the user to select the WIM size according to the accurazy required for accomplishing the task.