# a-sixel
A-Sixel library for encoding sixel images.
#### Basic Usage
```rust
use a_sixel::ADUSixelEncoder;
use image::RgbImage;
let img = RgbImage::new(100, 100);
println!("{}", <ADUSixelEncoder>::encode(&img));
```
### Choosing an Encoder
- I want fast encoding with good quality:
- Use [`ADUSixelEncoder`]
- [`BitSixelEncoder`] can also produce pretty good results for 256 colors
depending on the image while being over 10x faster.
- I'm time constrained:
- Use [`ADUSixelEncoder`] or [`BitSixelEncoder`]. You can customize `ADU`
by lowering the `STEPS` parameter to run faster if necessary while still
getting good results.
- I'm _really_ time constrained and can sacrifice a little quality:
- Use [`BitSixelEncoder<NoDither>`].
- I want high quality encoding, and don't mind a bit more computation:
- Use [`FocalSixelEncoder`].
- This matters a lot less if you're not crunching the palette down below
256 colors.
- Note that this an experimental encoder. It will *likely* produce better
results than just [`ADUSixelEncoder`], but it may not always do so. On
the test images, for my personal preferences, I think it's slightly
better - particularly at small palette sizes.
<details>
<summary>How it works</summary>
Under the hood, it is a modified version of the `ADUSixelEncoder`
that uses a weighted selection algorithm for its sample pixels. These
weights are determined based on saliency maps and measures of
statistical noise in the image.
In addition to the weighted selection, the distance metric used to
determine which cluster to place a pixel into also incorporates the
weight. Similar pixels with different weights will be nudged towards
clusters with similar weights. This is a mild effect, but it seems
to improve things over basic clustering when there are a lot of similar
colors in an image.
</details>
License: MIT OR Apache-2.0