d3.js 3D array interpolation

Question

Code is here: http://jsfiddle.net/S48QX/.

I want to draw a image based on a 3D data set, for example:

var data = [
    {x:1.428, y:0.500, energy:0.458},
    {x:1.428, y:1.191, energy:0.616},
    {x:1.428, y:1.882, energy:0.795},
    {x:1.428, y:2.573, energy:0.642},
    {x:1.428, y:3.264, energy:0.536},
    {x:1.428, y:3.955, energy:0.498},
    {x:1.428, y:4.646, energy:0.494},
    {x:1.428, y:5.337, energy:0.517},
     ...

}

It's like scattered plot, but I need every pixel to be set, not just a bunch of color dots on the image. So, my question is how can I interpolate scattered dots with d3.js.

The generated image here is the best I can do so far, but is it possible to make it more smooth and beautiful?

I am seeking a way to generate a HEATMAP only based on partial/scattered data. I hope there is a way in d3.js that can interpolate the missing part.

(1,5) ?   ?   ?  (5,5)
  ?   ?   ?   ?    ?
  ?   ?   ?   ?    ?
(1,2) ?   ?   ?  (5,2)

Answer 1

Plotly.js is based on D3.js and able to create contour plots for scattered data:

https://jsfiddle.net/vwksaob3/

var data = [ {      
  x: [0, 1, 1, 0],
  y: [0, 0, 1, 1],
  z: [0, 0, 1, 1],
  type: 'contour',  
  colorscale: 'Jet',
  showscale: false,
  autocontour: true  
}];

var layout = {
margin: {
b: 0,
l: 0,
r: 0,
t: 0
},
height: 600,
width: 600,
  title: '',
  xaxis: {
        ticks: '',
      showticklabels: false  

  },
  yaxis: {
       ticks: '',
       showticklabels: false     
  } 
};

Plotly.newPlot('graph', data, layout, {displayModeBar: false});

Answer 2

With some effort you might be able to translate an existing algorithm to JavaScript.

• Octave is open source and provides a method for scattered data interpolation:

http://www.dm.unibo.it/~achilles/calc/octave.html/Interpolation-on-Scattered-Data.html

The source code of Octave is available at

ftp://ftp.gnu.org/gnu/octave/

The file griddata.m and some referenced files can be found in the folder

octave_{version}\scripts\geometry

• D3.js seems to provide some voronoi and delaunay functionality that might be helpful:

https://github.com/d3/d3/wiki/Voronoi-Geom

http://bl.ocks.org/mbostock/4341156

• Python also provides a griddata method:

http://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.interpolate.griddata.html#scipy.interpolate.griddata

Answer 3

I have one solution using svg filters. Be careful as this may not be what you want since the mathematical interpretation of this interpolation would be more 'blur'. I mostly did it as an exercise on svg filters. However 2d interpolation end up with similar results: see cubic interpolation for example (http://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.interpolate.interp2d.html in python)

I used circles (you could try with rectangles) slightly overlapping and semi transparent and applied to gaussian blur on them, resulting in a 'heatmap' looking thing.

var filter = svg.append("defs").append('filter')
    .attr('id', 'blur')
    .append("feGaussianBlur")
    .attr("stdDeviation", 8);

then using .style('fill-opacity', 0.5).attr("filter", "url(#blur)") on the circles

See the fork http://jsfiddle.net/eqt1mkov/