This paper discusses the problem of inserting 3D models into a single image. The main focus of the paper is on the accurate recovery of the camera's parameters, so that 3D models can be inserted in the “correct” position and orientation. The paper addresses two issues. The first is an automatic extraction of the principal vanishing points from an image. The second is a theoretical and an experimental analysis of the errors. To test the concept, a system that “plants” virtual 3D objects in the image was implemented. It was tested on many indoor augmented-reality scenes. Our analysis and experiments have shown that errors in the placement of the objects are unnoticeable.