Mesh Match features are implemented through C++ plugins and python commands wrapped in a user friendly UI.
The core registration process works by first computing unique shape descriptors for points across both meshes. An iterative solver then compares these descriptors at global and local levels to establish correspondence, progressively aligning the meshes while minimizing deformation energy to ensure smooth, natural-looking results.
Complementing alignment, robust retargeting features allow the transfer of deformations between meshes—even with differing topologies—accurately accounting for local curvature and scale variations. Transformation matrices can also be retargeted, enabling, for example, the transfer of complete rig joint positions along with mesh deformations.
Performance is key: the plugin employs extensive multithreading, intelligent caching to avoid redundant calculations when attributes change, and optimized computation paths for symmetrical meshes.
Registration:
- Initial Alignment: Multiple options are available to pre-align the target mesh to the reference, utilizing centroid and bounding box data.
- Custom Landmarks (“Pins”) – Optional: Users can introduce landmark pairs to define known corresponding points, guiding the registration. Built-in weighting, color visualization, and adjustment features enhance control during pre-alignment, main registration, and post-refinement. (A companion MeshMatchLoc plugin automatically draws interactive locators for easy landmark selection and editing).
The pin pairs are way more than just points in space, it’s a complete deformer system on its own with many weighting options and the ability to layer the deformations. - Main Registration: The core solver iteratively identifies similar regions between the meshes based on shape descriptors, progressively aligning them while carefully controlling deformation to maintain overall rigidity and prevent overshooting.
- Post Refinement: After the main registration achieves a close match, a custom projection and relaxation algorithm further refines the target mesh’s alignment, ensuring it precisely overlaps the reference surface. For symmetrical meshes, a mirror axis option optimizes computation by processing only half the points.
Retargeting (same or different topologies):
Once a successful registration map has been computed (linking points or regions between the reference and the now-aligned target mesh), Mesh Match enables data retargeting. This process transfers information from the reference mesh back onto the original, undeformed topology of the target mesh.
This capability is crucial and works effectively regardless of whether the meshes share the same or different topologies.
Key retargeting functionalities include:
- Deformation Transfer: Propagate complex deformations like blend shape targets or skin weights from the reference mesh onto the target mesh’s structure.
- Matrix Transfer: Retarget transformation matrices. This is particularly useful for transferring skeletal rig joint positions based on the mesh alignment, effectively mapping a rig from one character shape to another.
- Attribute & Feature Transfer: Transfer per-vertex data such as UV coordinates, vertex colors, or other custom attributes present on the reference mesh to the corresponding locations on the target mesh.
Intelligent Transfer: The retargeting algorithms are designed to intelligently adapt the transferred data. They account for variations in local surface curvature and scale between the reference and target meshes, ensuring the data is propagated accurately and preserves the intended details on the target structure.
Notes:
Performance varies with mesh density—from near real-time editing for meshes with 2,000–4,000 points to a few seconds for higher-resolution meshes. Although not designed for live animated scenarios, the primary focus remains on achieving high-quality, production-ready results, with the entire setup process typically completed within minutes.