Rust library for converting RGB colors to the CIE-L*a*b* color space

extern crate lab;
use lab::Lab;

let pink_in_lab = Lab::from_rgb(&[253, 120, 138]);
// Lab { l: 66.639084, a: 52.251457, b: 14.860654 }

extern crate lab;
extern crate image;

use lab::Lab;
use image::Rgba;

let pixel: Rgba<u8> = Rgba { data: [253, 120, 138, 255] };
let lab = Lab::from_rgba(&pixel.data);
// Lab { l: 66.639084, a: 52.251457, b: 14.860654 }

Experimental SIMD functions

The lab::simd module is compiled for the x86_64 cpu architecture. If the current cpu can run AVX and SSE 4.1 operations, it can make use of the exported functions.

extern crate lab;
use lab::Lab;
#[cfg(target_arch = "x86_64")]
use lab::simd;

fn convert_rgbs(rgbs: &[[u8; 3]]) -> Vec<Lab> {
  // It's boilerplate, but it's also experimental. So.
  #[cfg(target_arch = "x86_64")]
  {
      if is_x86_feature_detected!("avx") && is_x86_feature_detected!("sse4.1") {
          return simd::rgbs_to_labs(rgbs);
      }
  }
  rgbs.iter().map(Lab::from_rgb).collect()
}

Performance increase over a serial map is wildly variable across CPUs, suggesting that there are still some optimizations to perform. A 2013 Macbook Air sees a ~25% decrease in benchmark times converting Labs to RGBs, and a ~40% decrease converting RGBs to Labs. Meanwhile a 6-core desktop computer sees near perfect 8x speedup converting Labs to RGBs, but a <10% improvement converting RGBs to Labs. Clearly it is a work in progress.

Minimum Rust version

Lab 0.7.0 requires Rust >= 1.31.0 for the chunks_exact slice method

Lab 0.6.0 can build as far back as Rust 1.13.0. Testing releases gets pretty tedious earlier than that.