Shape representation by metric interpolation

摘要: Coordinates of vertices in a triangulated surface can be efficiently represented as set coefficients that multiply given basis functions. One such natural orthonormal is provided by the eigenfunctions Laplace-Beltrami operator shape. The this case are nothing but result scalar inner product coordinates treated smooth function on shape and form basis. Keeping only significant allows for efficient representation under practical transformations. Selecting regular metric construction we notice while general preserved, important fine details often washed out. At other end, using scale invariant defining corresponding basis, preserves at potential expense loosing structure Here, adopt best both worlds. By finding right mix between one select serves representation-basis generator We use mean square error (MSE) to optimal space representation, compare results classical spectral techniques.