A Dynamic 3D, Object-Generating Algorithm from Multiple Viewpoints Using the Volume Intersection and Stereo Matching Methods

We are researching a modeling algorithm for the moving human body to generate a fine, dynamic, 3D object. In this paper, we propose a modeling method from multiple viewpoints. This modeling method uses an approximate shape, obtained by the volume intersection method with restrictive conditions set by the stereo matching method.In addition, in order to map the image texture on the surface of the dynamic 3D objects, we propose a vertex texture mapping method based on ray-casting. Our modeling experiments were performed using 19 Firewire cameras. We were able to confirm that the proposed methods are effective for generating fine, dynamic, 3D objects. The dynamic 3D object generated can be displayed as an “animation” by using a VRML browser.