HDRColorTypes

This package is an add-on to ColorTypes, and provides color types for high dynamic range imaging.

Note that types such as RGB{Float16} are already supported in ColorTypes. Currently this package supports the Radiance formats RGBE32 and XYZE32.

RGBE32 and XYZE32

RGBE32 and XYZE32 are color types based on floating-point numbers with three color component mantissa parts and one exponent part shared by the three. The mantissa parts and exponent part are all represented in 8-bit, and the colors have the total size of 32-bit.

RGBE32 and XYZE32 are defined as types with four fields of type UInt8 as follows.

struct RGBE32{T<:AbstractFloat} <: AbstractRGBE32{T}
    c1::UInt8
    c2::UInt8
    c3::UInt8
    e::UInt8
end

struct XYZE32{T<:AbstractFloat} <: AbstractXYZE32{T}
    c1::UInt8
    c2::UInt8
    c3::UInt8
    e::UInt8
end

The fact that they have four fields contrasts with the fact that ARGB32 and RGB24 have one field of type UInt32. Also, the field names are not r, g, b or x, y, z, but c1, c2, c3. This is to prevent direct access to the fields by mistaking a color for RGB or XYZ, in addition to the viewpoint of generality.

As the definitions above indicate, RGBE32 and XYZ32 are parametric types with "component type" T. This is common with RGB{T} and XYZ{T} and different from ARGB32 and RGB24. If T is not explicitly specified in the constructor, Float32 is applied.

julia> using HDRColorTypes
julia> using FixedPointNumbers
julia> RGBE32(100, 0, 0) # constructs an `RGBE32{Float32}`RGBE32{Float32}(100.0f0,0.0f0,0.0f0)
julia> RGB(100.0, 0, 0), RGB(1, 0, 0) # The determination of `T` differs from `RGB`.(RGB{Float64}(100.0,0.0,0.0), RGB{N0f8}(1.0,0.0,0.0))
julia> RGBE32{Float16}(10, 20, 30) # You can specify `T` explicitly.RGBE32{Float16}(Float16(10.0),Float16(20.0),Float16(30.0))
julia> XYZE32{Float64}(10, 20, 30)XYZE32{Float64}(10.0,20.0,30.0)

Of course, the type of the component values is T, and the values are lazily converted.

julia> using ColorTypes
julia> red(RGBE32{Float16}(3.14, 0, 0))Float16(3.14)
julia> comp2(XYZE32{Float32}(0, 2.71, 0)) # The actual mantissa part is only 8-bit2.703125f0

Note that regardless of the component type T, the entities of the instances of RGBE32 and XYZE32 are 32-bit.

julia> sizeof(RGBE32{Float32})4
julia> sizeof(RGBE32{Float64})4

To be compatible with the RGB and XYZ constructors, it is not allowed to pass directly the mantissa and exponent parts to the constructor arguments. You can use the factory functions rgbe32 and xyze32, and their generic version ccce32.

julia> rgbe32(0x12, 0x34, 0x56, 0x78)RGBE32{Float32}(0.0002746582f0,0.00079345703f0,0.0013122559f0)
julia> xyze32(0x12345678)XYZE32{Float32}(0.0002746582f0,0.00079345703f0,0.0013122559f0)
julia> ccce32(RGBE32{Float64}, 0x12345678)RGBE32{Float64}(0.000274658203125,0.00079345703125,0.001312255859375)

julia> reinterpret(UInt32, rgbe32(0x12345678))0x78563412