Getting started

ColorQuantization.jl can be applied to any iterable collection of Colorants img, e.g. an image or any AbstractArray. We demonstrate ColorQuantization.jl on the peppers test image:

using ColorQuantization
using ColorTypes
using TestImages

img = testimage("peppers")

To extract a color palette from an image, call quantize with a quantization method of your choice.

Clustering

Clustering methods such as KMeansQuantization are convenient since they allow us to specify the exact number of output colors:

quantize(img, KMeansQuantization(16))

quantize(img, KMeansQuantization(64))

quantize(img, KMeansQuantization(256))

Uniform quantization

If perfomance is crucial, UniformQuantization can be used to quantize colors to the closest points on a equidistant 3D grid.

Note

In constrast to clustering, here the argument to UniformQuantization specifies the number of subdivisions of the $[0, 1]³$ RGB color cube. The amount of output colors is therefore bounded by n³.

quantize(img, UniformQuantization(4))

quantize(img, UniformQuantization(6))

quantize(img, UniformQuantization(8))

Using UniformQuantization, every single pixel in the input image is rounded and appears in the generated color palette, which explains the blues and purples from the shadows of the peppers.

Output color type

It is also possible to specify the output color type:

colors = quantize(HSV, img, UniformQuantization(4))

typeof(colors)

Vector{HSV{Float32}} (alias for Array{ColorTypes.HSV{Float32}, 1})

This page was generated using Literate.jl.