Follow the OG paper's prescription of the SVD threshold

It makes it less likely to have wild stuff happens on spiky geometry
JulyCode · Feb 23, 2024 · 913f879 · 913f879
1 parent 6e8fcba
commit 913f879
Showing 1 changed file with 3 additions and 2 deletions.
diff --git a/Isosurfacing_3/include/CGAL/Isosurfacing_3/internal/dual_contouring_functors.h b/Isosurfacing_3/include/CGAL/Isosurfacing_3/internal/dual_contouring_functors.h
@@ -167,8 +167,9 @@ bool cell_position_QEM(const typename Domain::Cell_descriptor& c,
 
   Eigen::JacobiSVD<typename Eigen_matrix_x::EigenType> svd(A, Eigen::ComputeThinU | Eigen::ComputeThinV);
 
-  // set threshold as in Peter Lindstrom's paper, "Out-of-Core Simplification of Large Polygonal Models"
-  svd.setThreshold(1e-3);
+  // Lindstrom's paper, "Out-of-Core Simplification of Large Polygonal Models": 1e-3
+  // Ju's paper, "Dual Contouring of Hermite Data": 1e-1
+  svd.setThreshold(1e-1);
 
   Eigen_vector_3 x_hat;
   x_hat << x_coord(com), y_coord(com), z_coord(com);