equivalent of ImageMagick.getimageproperty()

[bjarthur]

is there a way to do this with TiffImages.jl?

using ImageMagick
wand = MagickWand()
readimage(wand, "filename.tif")
my_custom_metadata_value = ImageMagick.getimageproperty(wand, "my_custom_metadata_key")

[tlnagy]

Could you give an example TIFF and/or key?

You can do something like

julia> img = TiffImages.load("/home/tlnagy/Downloads/example.tiff")
....
julia> img.ifds[1]
IFD, with tags: 
        Tag(IMAGEWIDTH, 1024)
        Tag(IMAGELENGTH, 768)
        Tag(BITSPERSAMPLE, UInt16[8, 8, 8, 8])
        Tag(COMPRESSION, COMPRESSION_NONE)
        Tag(PHOTOMETRIC, 2)
        Tag(FILLORDER, 1)
        Tag(STRIPOFFSETS, UInt32[8, 131080, 262152, 393224, 524296, ...])
        Tag(ORIENTATION, 1)
        Tag(SAMPLESPERPIXEL, 4)
        Tag(ROWSPERSTRIP, 32)
        Tag(STRIPBYTECOUNTS, UInt32[131072, 131072, 131072, 131072, 131072, ...])
        Tag(PLANARCONFIG, 1)
        Tag(RESOLUTIONUNIT, 2)
        Tag(EXTRASAMPLES, 1)
        Tag(SAMPLEFORMAT, UInt16[1, 1, 1, 1])
        Tag(ICCPROFILE, Any[0, 0, 2, 36, 97, ...])

Accessing existing tags:

julia> img.ifds[1][TiffImages.IMAGEWIDTH]
Tag(IMAGEWIDTH, 1024)

Creating new tags:

julia> img.ifds[1][TiffImages.IMAGEDESCRIPTION]
ERROR: KeyError: key 0x010e not found
Stacktrace:
 [1] getindex
   @ ~/.julia/packages/OrderedCollections/PRayh/src/ordered_dict.jl:380 [inlined]
 [2] getindex
   @ ~/.julia/dev/TiffImages/src/ifds.jl:30 [inlined]
 [3] getindex(ifd::TiffImages.IFD{UInt32}, key::TiffImages.TiffTag)
   @ TiffImages ~/.julia/dev/TiffImages/src/ifds.jl:29
 [4] top-level scope
   @ REPL[5]:1

julia> img.ifds[1][TiffImages.IMAGEDESCRIPTION] = "test"
"test"

And if using a non-standard tag, you can always refer to it by its UInt16 code. Say we make a new tag

julia> img.ifds[1][UInt16(5678)] = "test"
"test"

julia> img.ifds[1]
IFD, with tags: 
        ...other stuff...
        Tag(UNKNOWN(5678), "test")
        Tag(ICCPROFILE, Any[0, 0, 2, 36, 97, ...])

[tlnagy]

Also, the tag accessing API is not finalized so if you had any thoughts on how to improve the behavior, I'm open to suggestions!

[bjarthur]

excellent, thanks!

in my case, i had to take the additional step of printing the data field of the Tag:

julia> x = img.ifds[1][TiffImages.IMAGEDESCRIPTION]
Tag(IMAGEDESCRIPTION, "<MetaData>
<prop id=...")

julia> print(x.data)
<MetaData>
<prop id="Description" type="string" value="Acquired from Hamamatsu DCAM Camera&#13;&#10;Exposure: 5.5 msec&#13;&#10;Binning: 1 X 1&#13;&#10;Region: 1024 x 1024, offset at (512, 512)&#13;&#10;Sensor Mode: Normal&#13;&#10;Digitizer: High-speed&#13;&#10;Bit Depth: 16-bit&#13;&#10;Frames to Average: 1&#13;&#10;Electron Count Conversion Factor: 0.48&#13;&#10;Electron Count Conversion Offset: 100&#13;&#10;"/>
<prop id="MetaDataVersion" type="float" value="1"/>
<prop id="ApplicationName" type="string" value="MetaMorph"/>
<prop id="ApplicationVersion" type="string" value="7.10.4.407"/>
<PlaneInfo>
<prop id="plane-type" type="string" value="plane"/>
<prop id="pixel-size-x" type="int" value="1024"/>
<prop id="pixel-size-y" type="int" value="1024"/>
...

[tlnagy]

Yeah, currently I truncate the output when displaying long strings to just the first 20 bytes here:

TiffImages.jl/src/tags.jl

Line 116 in 18e6b45

_showdata(io::IO, t::Tag{<: AbstractString}) = print(io, "\"", first(t.data, 20), (length(t.data) > 20) ? "..." : "", "\"")

But I'm happy to consider alternative approaches.

[bjarthur]

re-opening with a suggestion: would be nice to get the tags as described above without having to read in the entire image. mmap is the suggested work-around i suppose? an API that provides a means to get the image size, bit depth, etc. would be better.

[tlnagy]

Mmap would certainly work in a pinch:

TiffImages.load("path/to/file.tiff"; mmap = true)

but it should be possible to load the tag info separately from the image data. I already load the IFD tag info separately at the beginning:

TiffImages.jl/src/load.jl

Lines 12 to 16 in 18e6b45

	for ifd in tf
	load!(tf, ifd)
	push!(ifds, ifd)
	nplanes += 1
	end

[bjarthur]

i'm not familiar enough with the internals of TIFF, but for an API, at a minimum, a method which returned the image dimensions given a filename would help. specifically, in my case each plane is stored in a separate file. so i need to preallocate a 3D array, and then loop over loading each image. and to do that i need the dimensions.

[tlnagy]

What if you could access all the ifd info as in #59 (comment), but without loading the image?

We could call that function

TiffImages.loadtags()

or

TiffImages.loadifds()

Not sure which one I prefer. loadifds is more accurate, but perhaps overly specific?

[bjarthur]

that'd be great! i'm not a TIFF expert so have no preference on your proposed alternatives.