This bl.ock takes an alternative approach to this one, in that it is not dependent on a circular projection disc.
It runs each pixel through a forward then inverse projection to see if the pixel should be plotted. It is in response to a SO question, [this was a potential answer] (http://stackoverflow.com/a/41856996/7106086). This is required in transforming the equiratangular source image as the d3 projection function is not one to one.
https://d3js.org/d3.v3.min.js
https://d3js.org/d3.geo.projection.v0.min.js