2D natural neighbor interpolation (nni)

For Node and browsers.

Some info about nni.
The nii method is used to interpolate sparse data point along a regular 2D grid surface. This methods work with Voronoi cells and area stilling and is said to give results that are close to the "real solution". The only issue with this method is that it's very computational, and thus, quite slow.

On the left: nni, on the right: same but with using some threshold at rendering.

Concept

Sparse point interpolations need:

• a set of sparse points. Here, we will call them seeds
• an output size (width and height)

This implementation performs the interpolation in two steps:

• build the interpolation map for each output pixel, aka the weight of each seeds involved in each output pixel
• use the interpolation map to create the output image

The purpose of creating the map at an intermediary step is to be able to reuse it for the same configuration but with different seed values, because what takes a while to generate is this interpolation map.

Install

npm install --save natninter

Use it in a HTML file:

<script src="natinter/dist/natninter.js"></script>

Use it in ES6:

import * as natninter from 'natninter';

Use it in Node:

const natninter = require("natninter");

Basic usage

// a list of seeds
var seeds = [
  {x: 100, y: 100,  value: Math.random()*255 },
  {x: 300, y: 100,  value: Math.random()*255  },
  {x: 300, y: 300, value: Math.random()*255 },
  {x: 100, y: 300, value: Math.random()*255  },
  {x: 200 , y: 200 , value: Math.random()*255 }
]

// creating the nni interpolator instance
var nnInter = new natninter.Interpolator();

// setting the output size
nnInter.setOutputSize(256, 256);

// add a list of seeds
nnInter.addSeeds( seeds );

// Create the interpolation map
// (this may take some seconds, start with a small image to benchmark it)
nnInter.generateMap();

// generate the output image witha  nice interpolation
var output = nnInter.generateImage();

The output is actually an object of the form:

{
  _data: Float32Array,
  _metadata: {
    width: Number,
    heigth: Number
  }
}

Where _data is a 1D array of length heightxwidth arrange in a row major way.

Examples

A regular example and A wavy example
See the folder examples for a Node and a browser example.

Generate a map and save it locally

natninter comes with a node executable to generate a samplig-map file in JSON format. In a terminal, find the executable generatemap.js, located in the bin folder. Then:

./generatemap.js --seeds=your/own/seedsfile.json -width=400 --height=400 --output=somehwere/map.json

You can use the seed file in bin/samples/seeds1.json and use it as a a model to write your own seed file.

Special thanks

Since I didn't want to reinvent the wheel and that I trust some good developers out there, here are the dependencies natninter uses:

• Compute Voronoi diagram - damn fast!
• Polygon area computation - super simple a pretty fast too!