A new method for capturing and reconstructing 3D representations of real objects in a virtual reality system is introduced. Virtual reality applications allow users to navigate and interact with the 3D objects through the environment. This interaction requires that the 3D representation of real objects be highly accurate in modeling the reality. The novelty of the new methodology proposed, consists on the fact that it uses only a high resolution (7 megapixels or higher) digital camera and a projector in conjunction with 3D surface reconstruction techniques based on non-uniform rational Bzier spline (NURBS) functions. The 3D object reconstruction is based on finding unique control points on the 2D images of the object and constructing corresponding 2-D NURBS curves which contain the control points through a process of NURNS fitting. The control points are situated on grid lines which are extracted from the object surface on which a color coded grid is projected. The 2-D NURBS curves are projected into a 3-D space to eventually re-create the 3-D surface of interest. The method does not require any a priori knowledge of the absolute positioning or orientation of the camera and the projector as other 3D reconstruction techniques do. The precision of the method depends on the camera resolution and can attain easily sub-millimeters ranges. Examples illustrate the process.