Blender Addon: Boundaries to Ngon

This tool provides a one-click solution to fill non-intersecting edge loops with a small set of ngons. It extends Blender's built-in Fill tool by correctly handling holes and nested polygons. Since the tool also accepts faces as input, it serves as an advanced "Dissolve Faces" tool.

Tested for Blender 3.5+ on Windows. In theory it should work on Mac (M1+), but I haven't tested it myself.

For other system architectures, the Addon falls back to a numpy-based implementation of delauny triangulation. This is indicated in the "Preferences"-Panel. This fallback may cause some glitchy Ngons in case of challenging geometry.

How to use:

1. Enter Edit Mode.
2. Select the boundary edges you want to fill.
3. Optionally select faces to include their boundary edges (edges with exactly one adjacent selected face will be filled).
4. Click "Create Ngons".

Key Features

Smart hole handling: Correctly processes holes by connecting them to outer polygons, creating 2+ ngons. Works with nested polygons ("holes within holes") and complex cases with hundreds of input edges.

Advanced face dissolution: Since you can input faces, this serves as an enhanced "Dissolve Faces" tool that properly handles holes.

Optimized ngon count: The algorithm minimizes the number of output ngons (typically just 2 ngons for standard cases). Complex scenarios may produce significantly more due to the computational complexity of this optimization problem.

Point weld support: Handles cases where two polygons touch at exactly one vertex, treating them as external polygons or holes based on their relative position.

Flatten option: Projects inner vertices to be roughly coplanar with outer vertices.

UV preservation: Maintains UV coordinates when desired, using projection matrices estimated from input faces.

Best Practices

To work properly,

• each edge must belong to exactly one closed loop, and
• no two edges should intersect on the projection plane

The algorithm projects 3D points to a 2D plane, working best when edges are roughly coplanar. Four projection methods are available to handle difficult cases:

• Simple: Assumes full coplanarity; determines projection plane from first 3 selected vertices
• Linear Approximation: Estimates plane from all selected vertices using SVD
• Active Face: Uses the active face's plane for projection
• From Object: Averages all faces from a selected mesh object to construct the projection plane

Note: Incorrect projections may cause edge intersections, resulting in degenerate ngons.

Performance

The algorithm is mostly written in python, which is fine for a few hundred edges as input. For very large dataset, expect poor performance. You can adjust the maximum number of iterations used by the algorithm in the preferences.