FastHistograms
FastHistograms declares and implements a minimal histogram interface with a focus on speed. This package does not aim to implement general histogram algorithms that work for a wide array of data types; for that purpose, consider StatsBase.
Currently, the only implemented algorithm is for fixed-width bins on (small) 2D data.
Example
using FastHistograms, Random
julia> h = SingleThreadFixedWidth2DHistogram()
julia> img1 = zeros(10, 10)
julia> img1[:, 6:end] .= 0xff
julia> img2 = zeros(10, 10)
julia> img2[6:end, :] .= 0xff
# Compute the histogram bin counts
julia> calc_hist!(h, img1, img2)
julia> @show counts(h) # Get the bin counts
16×16 Matrix{Int64}:
25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 25
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 25Benchmarks
With two 40x80 8-bit images and 16 bins per dimension, FastHistograms runs in 12 μs and StatsBase runs in 194 μs.