Restore consistency with old color distances

The goal ultimately is that `abs2(g::Gray) == abs2(RGB(g::Gray))`, but
for this package this will be a breaking change. Consquently, for now
let's measure distances in a manner consistent with older
ColorVectorSpace releases.

Author: timholy

src/core.jl

@@ -7,7 +7,9 @@ accum_type(::Type{C}) where {C<:Colorant}   = base_colorant_type(C){accum_type(e
 accum_type(val) = isa(val, Type) ? throw_accum_type(val) : convert(accum_type(typeof(val)), val)
 throw_accum_type(T) = error("type $T not supported in `accum_type`")
 
-_abs2(c::MathTypes) = c ⋅ c
+# TODO:
+# _abs2(c::MathTypes) = c ⋅ c
+_abs2(c::MathTypes) = mapreducec(v->float(v)^2, +, 0, c)
 _abs2(x) = abs2(x)
 
 default_diff_fn(c1::CT1,c2::CT2) where {CT1<:Union{Colorant,Real}, CT2<:Union{Colorant,Real}} = sqrt(_abs2(c1-accum_type(c2)))

test/flood_fill.jl

@@ -56,7 +56,7 @@ using Test
     # Colors
     path = download("https://github.com/JuliaImages/juliaimages.github.io/raw/source/docs/src/pkgs/segmentation/assets/flower.jpg")
     img = load(path)
-    seg = flood(img, CartesianIndex(87,280); thresh=0.3)
+    seg = flood(img, CartesianIndex(87,280); thresh=0.3*sqrt(3))   # TODO: eliminate the sqrt(3) when we transition to `abs2(c) = c ⋅ c`
     @test 0.2*length(seg) <= sum(seg) <= 0.25*length(seg)
     c = mean(img[seg])
     # N0f8 makes for easier approximate testing