/surface-nets

Arbitrary dimensional level sets

Primary LanguageJavaScriptMIT LicenseMIT

surface-nets

Extract a simplicial level set from an ndarray in any dimension using naive surface nets. This module works in both node.js and with browserify!

If you are doing experiments with greedy meshing or working with rectangular data, then you might want contour2d instead.

Example

Here is a 2D example:

//Load modules
var surfaceNets = require("surface-nets")
var ndarray = require("ndarray")
var fill = require("ndarray-fill")

//Initialize array to a circle
var array = ndarray(new Float32Array(32*32), [32,32])
fill(array, function(i,j) {
  return Math.pow(i-16,2) + Math.pow(j-16,2)
})

//Extract 2D contour (this is all there is to it!)
var complex = surfaceNets(array, 15*15)

//Write SVG image to stdout
var svgFile = ['<svg xmlns="http://www.w3.org/2000/svg" width="320" height="320">']
complex.cells.forEach(function(cell) {
  var p0 = complex.positions[cell[0]]
  var p1 = complex.positions[cell[1]]
  svgFile.push('<line x1="', 10*p0[0], '" y1="', 10*p0[1], '" x2="', 10*p1[0], '" y2="', 10*p1[1], '" stroke="red" stroke-width="1" />')
})
complex.positions.forEach(function(p) {
  svgFile.push('<circle cx="', 10*p[0], '" cy="', 10*p[1], '" r="1" stroke="black" stroke-width="0.1" fill="black" />')
})
svgFile.push('</svg>')
console.log(svgFile.join(""))

And here is the output SVG:

This module also works in 3D. Here is an example:

//Load modules
var surfaceNets = require("surface-nets")
var ndarray = require("ndarray")
var fill = require("ndarray-fill")
var mat4 = require("gl-matrix").mat4

//Initialize array
var array = ndarray(new Float32Array(32*32*32), [32,32,32])
fill(array, function(i,j,k) {
  return Math.pow(i-16,2) + Math.pow(j-16,2) + Math.pow(k-16,2)
})

//Generate surface! (again, just one line)
var complex = surfaceNets(array, 100)

//Render the implicit surface to stdout
console.log('<svg xmlns="http://www.w3.org/2000/svg" width="512" height="512" version="1.1">')
console.log(require("svg-3d-simplicial-complex")(
  complex.cells, 
  complex.positions, {
    view: mat4.lookAt(
      mat4.create(), 
      [32, 32, 32], 
      [16, 16, 16], 
      [0,1,0]),
    projection: mat4.perspective(mat4.create(),
      Math.PI/4.0,
      1.0,
      0.1,
      1000.0),
    viewport: [[0,0], [512,512]]
  }))
console.log("</svg>")

And here is the result:

And while it is a bit trivial, you can also generate surfaces in 1D:

var surfaceNets = require("surface-nets")
var ndarray = require("ndarray")

console.log(surfaceNets(ndarray([1, -1, 0, 5, -10])))

Output:

{ positions: [ [ 0.5 ], [ 2 ], [ 3.3333333333333335 ] ],
  cells: [ [ 0 ], [ 1 ], [ 2 ] ] }

The code should work in 4D and higher dimensions, but this is not well tested and it is harder to visualize. (Also, why would you want to bother!?!)

Install

npm install surface-nets

API

require("surface-nets")(array[,level])

Extracts the level set at level from array as a simplicial complex.

  • array is an ndarray
  • level is an optional number which determines the level at which the levelset is evaluated (default 0)

Returns An object with a pair of properties representing a simplicial complex:

  • positions is an array encoding the positions of the vertices. The coordinates of the positions are with respect to the indices in array.
  • cells is an array encoding the cells of the simplicial complex as tuples of indices into the position array.

Credits

(c) 2014 Mikola Lysenko. MIT License