PoissonRecon-rs

Safe Rust bindings for the Poisson Surface Reconstruction algorithm implemented by Michael Kazhdan.

Original source: https://github.com/mkazhdan/PoissonRecon

Example

let mesh = poissonrecon::reconstruct_surface(
    &points,
    &normals,
    &poissonrecon::PoissonParamsBuilder::default().build(),
)?;

Comparison to the poisson_reconstruction crate

The poisson_reconstruction crate is an excellent Rust reimplementation of the Screened Poisson Surface Reconstruction algorithm. It's maintained by the Foresight Mining Software Corporation.

This poissonrecon crate on the other hand, is a Rust wrapper around Michael Kazhdan's original C++ implementation. It compiles a C++ library that you must drag along with your Rust project. It also exposes more tuning parameters for the algorithm than the poisson_reconstruction crate does, which may be useful for some applications.

You should use the poisson_reconstruction crate if you want a pure Rust implementation.

You should use the poissonrecon crate if you want to use the original battle-tested implementation, and you don't mind dragging along a C++ library.

The repository contains an example comparing the output of the two implementations:

cargo run --release --example comparison

Template unrolling

The PoissonRecon C++ algorithm is templated over the FEMDegree and the BoundaryType. You can enable compilation of different combinations of FEMDegree and BoundaryType by enabling or disabling the corresponding feature flags.

These are the available feature flags:

• degree_1: Enables the [Degree::One] option.
• degree_2: Enables the [Degree::Two] option.
• degree_3: Enables the [Degree::Three] option.
• boundary_free: Enables the [BoundaryType::Free] option.
• boundary_dirichlet: Enables the [BoundaryType::Dirichlet] option.
• boundary_neumann: Enables the [BoundaryType::Neumann] option.

degree_1 and boundary_neumann are enabled by default.

⚠️ You must have at least one degree and one boundary type enabled. If none are enabled, the library will forcefully compile as if degree_1 and boundary_neumann were enabled.

It is recommended to enable only the feature flags you need, as each combination will significantly increase the compilation time and binary size. Enabling all feature flags will result in a 3 * 3 = 9 increase in compilation time and binary size.