fix: add square root back to distance fn

src/interpolation/mod.rs

@@ -32,9 +32,9 @@ pub trait InterpolatedRemapper<'a>: Sync {
 }
 
 type ColorTree = KdTree<f64, u16, 3, 4, u16>;
-fn squared_euclidean(a: &[f64; 3], b: &[f64; 3]) -> f64 {
+fn euclidean(a: &[f64; 3], b: &[f64; 3]) -> f64 {
     let dl = (a[0] - b[0]).powi(2);
     let da = (a[1] - b[1]).powi(2);
     let db = (a[2] - b[2]).powi(2);
-    dl + da + db
+    (dl + da + db).sqrt()
 }

src/interpolation/nearest_neighbor.rs

@@ -1,7 +1,7 @@
 use image::Rgb;
 use oklab::{srgb_to_oklab, Oklab};
 
-use super::{squared_euclidean, ColorTree, InterpolatedRemapper};
+use super::{euclidean, ColorTree, InterpolatedRemapper};
 use crate::{GenerateLut, Image};
 
 /// Simple remapper that doesn't do any interpolation. Mostly used internally by the other
@@ -40,7 +40,7 @@ impl<'a> InterpolatedRemapper<'a> for NearestNeighborRemapper<'a> {
         let Oklab { l, a, b } = srgb_to_oklab(pixel.0.into());
         let (_, nearest) = self.tree.nearest_one(
             &[l as f64 * self.lum_factor, a as f64, b as f64],
-            &squared_euclidean,
+            &euclidean,
         );
         *pixel = Rgb(self.palette[nearest as usize]);
     }

src/interpolation/rbf/mod.rs

@@ -2,7 +2,7 @@ use std::f64;
 
 use kiddo::float::{kdtree::KdTree, neighbour::Neighbour};
 
-use super::{squared_euclidean, ColorTree, InterpolatedRemapper};
+use super::{euclidean, ColorTree, InterpolatedRemapper};
 use crate::GenerateLut;
 
 pub mod gaussian;
@@ -70,7 +70,7 @@ impl<'a, F: RadialBasisFn> InterpolatedRemapper<'a> for RBFRemapper<F> {
         match &self.tree {
             None => {
                 for p_color in self.palette.iter() {
-                    let distance = squared_euclidean(&color, p_color);
+                    let distance = euclidean(&color, p_color);
                     let weight = self.rbf.radial_basis(distance);
 
                     numerator[0] += p_color[0] * weight;
@@ -81,7 +81,7 @@ impl<'a, F: RadialBasisFn> InterpolatedRemapper<'a> for RBFRemapper<F> {
             },
             Some((nearest, tree)) => {
                 for Neighbour { item, distance } in
-                    tree.nearest_n(&color, *nearest, &squared_euclidean)
+                    tree.nearest_n(&color, *nearest, &euclidean)
                 {
                     let weight = self.rbf.radial_basis(distance);
                     let p_color = self.palette[item as usize];