Translate ndarrays with sinc interpolation
JavaScript
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
example
test
.gitignore
.travis.yml
LICENSE
README.md
package.json
translate.js

README.md

ndarray-translate-fft

Translates an array using sinc interpolation. For some things this makes sense, but for other signals this may not be what you want. (For example, you can get bigger values or negative stuff in your signal using this method). If you want to use bilinear interpolation, check out ndarray-warp, or if all your coordinates are integers try ndarray-translate instead.

build status

Example

Zero padded

Here is a simple example showing how to warp with 0-padding boundary conditions:

var baboon = require("luminance")(require('baboon-image'))
var translate = require("../translate.js")

translate(baboon, [100, 180])

require("save-pixels")(baboon, "png").pipe(process.stdout)

Output

Wrapped

ndarray-fft can also handle periodic boundary conditions by replacing the translate line with the following:

var baboon = require("luminance")(require('baboon-image'))
var translate = require("../translate.js")

translate.wrap(baboon, [100, 180])

require("save-pixels")(baboon, "png").pipe(process.stdout)

Output

Install

Install using npm:

npm install ndarray-translate-fft

API

var translate = require("ndarray-translate-fft")

translate(array, shift)

Translates array by shift amount in place using sinc interpolation with 0-boundary conditions.

  • array is an ndarray to translate (get mutated)
  • shift is an array of numbers indicated the amount to shift by (can be a fractional number), should have the same length as array.shape

Returns array

Note that due to not using an infinite amount of padding (and/or using additional corrections), the zero-padded result might not be completely accurate (but since in this case you are shifting out data anyway...).

translate.wrap(array, shift)

Translates an array by shift amount in place using periodic boundary conditions. This is exactly recoverable.

  • array is the array to translate
  • shift is the amount to shift by

Returns array

Reasons to use ndarray-translate-fft:

  • You want to translate by fractional amounts
  • You want your translations to be exactly invertible
  • You want to handle periodic boundary conditions

Reasons NOT to use ndarray-translate-fft

  • You are shifting with zero padding and integer vectors
  • You are concerned about speed.

License

(c) 2013 Mikola Lysenko. MIT License