ImageMesh: Low-Poly 3D Reconstruction

Author: Andrew R. Garcia

—

Introduction

ImageMesh is a method in VoxelMap for turning 2D images into 3D meshes. It works by slicing an image into sectors and building Convex Hulls in each sector, producing a lightweight .obj model.

  • Voxel-based (ImageMap) → dense voxel cloud

  • Mesh-based (ImageMesh) → compact polygonal mesh

ImageMesh is best when you want fast, low-poly 3D models suitable for Blender, PyVista, or game engines.

—

Quick Example

from voxelmap import Model
from voxelmap.mesh import ImageMesh, MeshView

model = Model(file="docs/img/land.png")

# create a low-poly mesh with sector partitioning
ImageMesh(model.array, out_file="land.obj", L_sectors=15)

# preview interactively (requires voxelmap[mesh])
MeshView("land.obj", flat_color="white", alpha=0.8)

➡ Produces a lightweight terrain mesh.

Low-poly mesh of terrain

—

When to Use ImageMesh

  • ✅ You want compact models (small .obj files).

  • ✅ You don’t need voxel-level detail.

  • ✅ You plan to edit meshes in Blender, Maya, etc.

  • ❌ If you need detailed volumetric rendering → use ImageMap instead.

—

Appendix: Technical Notes

ImageMesh partitions the input image into sectors. Each sector is mapped to 3D points by treating pixel intensity as depth, then enclosed by a Convex Hull.

  • The algorithm uses scipy.spatial.ConvexHull (QuickHull).

  • Increasing sectors captures more local detail (see figure below).

  • Mesh complexity is proportional to the number of sectors, not the number of pixels.

Effect of sectors

Left: 1 sector. Middle: 4 sectors. Right: 16 sectors.

Complexity (summary): - Time ≈ \(\mathcal{O}(w h n \log n)\) - Space ≈ \(\mathcal{O}(n \sqrt{s}/3)\)

Where \(w, h\) are image size, \(n\) points per sector, and \(s\) sectors.

—

Conclusion

ImageMesh gives you a fast, memory-efficient way to turn 2D images into usable 3D assets. For advanced users, the original complexity analysis and whitepaper are preserved here for reproducibility.