BITSPERSAMPLE gets overwritten from 8 to [16, 16, 16]

[lanceXwq]

This may be partially related to #72, but seems to cover only a specific aspect of it.

The attached file is expected to be an 8-bit grayscale image stack. However, when loaded using TiffImages.jl, it is interpreted as a 16-bit RGB image.

Upon investigating, I suspect the issue may stem from this line:

TiffImages.jl/src/load.jl

Line 71 in ff965a6

data = fixcolors(loaded, first(ifds))

Specifically, fixcolor appears to convert eltype(data) from TiffImages.Palette{FixedPointNumbers.N0f8} to ColorTypes.RGB{FixedPointNumbers.N0f16}.

As a result, this later line:

TiffImages.jl/src/types/dense.jl

Line 138 in ff965a6

ifd[BITSPERSAMPLE] = collect(UInt16.(bitspersample(data)))

overwrites the IFD's BITSPERSAMPLE tag from Tag(BITSPERSAMPLE, 8) to Tag(BITSPERSAMPLE, UInt16[16, 16, 16]).

I observed similar behavior with the TIF file referenced in #72.

Could you clarify whether this is the intended behavior? If not, I’d be happy to help fix it. 😃

ROI_2_camera_A_raw.tiff.zip

[tlnagy]

Thanks for the bug report @lanceXwq. My quick assessment is that bitspersample isn't properly handling the wrapped palette color image. There should probably be another function definition of bitspersample needed here:

TiffImages.jl/src/types/common.jl

Line 40 in ff965a6

bitspersample(img::AbstractArray) = bitspersample(eltype(img))

to handle the palette images. Happy to review a PR if you want to submit a fix!