Struct exoquant::Quantizer

pub struct Quantizer(_);

The main color quantizer state.

The Quantizer is used to find a palette of colors that represent the colors in an input Histogram as well as possible.

To use it you first create a new Quantizer instance using Quantizer::new, then call quantizer.step() until quantizer.num_colors() reaches your target color count, then call quantizer.colors() to get your final palette Vec of Colors.

Optionally you can call quantizer.optimize() between calls to quantizer.step() to run K-Means optimizations during quantization.

If you don't intend to do K-Means optimzations during quantization, you can also just use the shortcut function Quantizer::create_palette.

Examples

let mut quantizer = Quantizer::new(&histogram, &colorspace);
while quantizer.num_colors() < 256 {
  quantizer.step();
}
let palette = quantizer.colors(&colorspace);

let palette = Quantizer::create_palette(&histogram, &colorspace, 256);

Methods

impl Quantizer

fn new<T: ColorSpace>(histogram: &Histogram, colorspace: &T) -> Quantizer

Create a new Quantizer state for the given histogram.

fn create_palette<T: ColorSpace>(histogram: &Histogram, colorspace: &T, num_colors: usize) -> Vec<Color>

A shortcut function to directly create a palette from a histogram.

fn num_colors(&self) -> usize

Returns the current number of colors in this Quantizer state.

This starts off at 1 and increases by 1 for each call to quantizer.step().

fn step(&mut self)

Run one quantization step which increases the num_colors() by one.

fn colors<T: ColorSpace>(&self, colorspace: &T) -> Vec<Color>

Returns colors the current Quantizer state represents..

fn optimize(self, optimizer: &Optimizer, num_iterations: usize) -> Quantizer

Run a number of K-Means iteration on the current quantizer state.

This can improve the quality of the final palette by a certain amount, but note that this is a fairly expensive operation. You might find it sufficient to only run optimizations on the final palette, in which case optimizer.optimize_palette is slightly less expensive.

Examples

let optimizer = optimizer::KMeans;
let mut quantizer = Quantizer::new(&histogram, &colorspace);
while quantizer.num_colors() < 256 {
  quantizer.step();
  if quantizer.num_colors() % 32 == 0 {
    quantizer = quantizer.optimize(&optimizer, 4);
  }
}
let palette = quantizer.colors(&colorspace);