Reconstruction of 3D models from range images usually focuses on complex objects completely contained in the field of view. Using range images to reconstructing a whole environment is challenging because of many occlusions. The focus of this paper is the reconstruction of the corners and edges of partially occluded simple-shape objects like furniture pieces in a indoor scene. Little research has been done on reconstructing obscured surface parts. We introduce a new approach to detect and complete occlusions. The proposed approach is based on the good boundary and surface continuation and explore architectural constraints as well. Results on real images confirmed improvement of environment modelling and the perception of the scene objects.