next up previous
Next: Simplification Examples Up: Computing Error Bounds Previous: Computing Error Bounds

Error Propagation

We choose a simple scheme for propagation of error from one triangulation to the next. For the geometry and each scalar variable, we associate two variables with each triangular face. One variable is a current upper bound on the ``positive'' error accumulated, the other represents ``negative'' error. Given the accumulated errors on the faces surrounding a candidate vertex, we compute and add in the errors which are introduced through deleting the vertex. Figure 7 illustrates the process of computing an upper bound on the error along an introduced edge based on propagated error from the previous triangulation and introduced error caused by retriangulation.

  
Figure 7: Propagating error to a new triangulation



Dan Schikore
Fri Oct 4 13:30:14 EST 1996