QOI.jl - Implementation of the QOI (Quite OK Image) format

This Julia package contains a decoder and encoder for the QOI image format. The QOI format is very simple and can be faster than using PNG (see the benchmarks) at the cost of a slightly worse compression ratio.

The code here is based on the reference C implementation given in https://github.com/phoboslab/qoi.

This is the simplest API and likely the one that most should use. Simply, use the load/save API from FileIO.jl to load and save QOI images.

using FileIO
image = load("test.qoi")
save("test2.qoi", image)

• QOI.qoi_decode(f::Union{String, IO}) - Read an image in the QOI format from the file/IO f and return a matrix with RGB or RGBA colorants from ColorTypes.jl
• QOI.qoi_encode(f::Union{String, IO}, image::AbstractMatrix{<:Colorant}) - Write the image to the the file/IO f.

The QOI format is read in row-major order. This means that a transpose is required to create the matrices returned in the basic API. To avoid this, the following more advanced APIs exist:

• QOI.qoi_decode_raw(v::AbstractVector{UIt8}}) - Takes a vector of the bytes of an image in QOI format and returns the uncompressed vector of bytes from decoding.

• QOI.qoi_encode_raw(image::AbstractVecOrMat{UInt8}, header::QOI.QOIHeader) - Returns the bytes from compressing the bytes in image. The required header is defined as

  struct QOIHeader
    width::UInt32
    height::UInt32
    channels::QOI.QOIChannel # @enum
    colorspace::QOI.QOIColorSpace # @enum
  end

  where channels can be either QOI.QOI_RGB or QOI.QOI_RGBA and colorspace can be either QOI.QOI_SRGB or QOI.QOI_LINEAR

• QOI.qoi_encode_raw!(data::AbstractVecOrMat{UInt8}, image::AbstractVecOrMat{UInt8}, header::QOIHeader) - Same as above except writing the compressed data into data.

The benchmarks here compares the speed of encoding/decoding images with QOI.jl and PNGFiles.jl (which uses libpng) It is supposed to mimic https://qoiformat.org/benchmark/. The images used in the benchmark are taken from TestImages.jl, specifically, all the PNG images in https://testimages.juliaimages.org/stable/imagelist/. The benchmarks were run on Linux on a 12th Gen Intel(R) Core(TM) i9-12900K CPU. They can be repeated by running the benchmark/runbenchmarks.jl file with the associated environment.