Gaussian splats on surface meshes — gaussian

Create per-vertex Gaussian falloff maps centred on coordinates or vertices. Distances can be measured either in straight-line Euclidean space or along the mesh using geodesic paths (Dijkstra on edge lengths).

Usage

gaussian_splat(surface, center, sigma, use_geodesic = FALSE)

gaussian_splat_vertex(surface, vertex_idx, sigma, use_geodesic = FALSE)

gaussian_splat_multi(
  surface,
  centers,
  sigmas,
  weights = NULL,
  use_geodesic = FALSE
)

Arguments

surface: A SurfaceGeometry.
center: Numeric vector of length 3 giving the centre coordinate.
sigma: Positive numeric standard deviation of the Gaussian kernel.
use_geodesic: Logical; when TRUE use along-surface distances.
vertex_idx: Integer index (1-based) of the centre vertex.
centers: Numeric matrix with three columns; each row is a centre.
sigmas: Numeric vector of length 1 or nrow(centers).
weights: Optional numeric vector of length 1 or nrow(centers) used to weight each centre (defaults to 1).

Value

A NeuroSurface with values \(\exp(-d^2 / (2 * sigma^2))\) at every vertex. For gaussian_splat_multi the per-centre maps are summed (after weighting).

Examples

geom <- example_surface_geometry()
g1 <- gaussian_splat(geom, center = c(0, 0, 0), sigma = 1)
v_idx <- find_nearest_vertex(geom, c(0.1, 0, 0))
g2 <- gaussian_splat_vertex(geom, vertex_idx = v_idx, sigma = 1)
g_multi <- gaussian_splat_multi(
  geom,
  centers = rbind(c(0, 0, 0), c(1, 0, 0)),
  sigmas = c(0.5, 1)
)