@zyzle/image-kmeans
v2.0.0
Published
A WebAssembly module providing k-means clustering calculation for JS canvas images
Downloads
5
Readme
@zyzle/image-kmeans
Generate an array of colours from an image based on k-means clustering.
Version 2.x changes and usage
Version 2.0 comes with some big changes to the way this module is used. The process to run the code looks something like the following:
import * as wasm from "@zyzle/image-kmeans";
const canvas = document.createElement("canvas");
const ctx = canvas.getContext("2d");
// draw an image to the canvas
// Instantiate the class passing in your 2d rendering context and
// the image width and height
const wasmInstance = new ImageKmeans(ctx, ibm.width, ibm.height);
After instantiating the class you now have 2 choices
Fixed K number of clusters
Use a pre-determined number of clusters for the calculations in this case 4:
const result = wasmInstance.with_fixed_k_number(4);
Derived K number
The module will do multiple runs of the k-means algorithm and determine the best fit for the number of selected clusters.
Note: Because this will perform multiple complete runs of the algorithm it may take significantly longer than a single run with a fixed number, although should give better results, most of the time this will return with 40s even for multi-megabyte images
const result = wasmInstance.with_derived_k_number();
Results object
Both of the above now return a RunResult object which looks like the following:
RunResult {
ks: number; // the number of k clusters used in this run
clusters: Array<Color>; // An array containing color objects
// { r: number, g: number, b: number }
// representing the cluster centroids
wcss: number // the combined within-cluster sum of squares
// for these clusters
}
Building form source
Use wasm-pack
to build the Rust source into WebAssembly, this will output the JS/Wasm into a pkg
folder using:
wasm-pack build
License
Licensed under MIT license (LICENSE or http://opensource.org/licenses/MIT)