Inferring 3D Volumetric Shape of Both Moving Objects and Static Background Observed by a Moving Camera