ImageJ deletes the exif from images

Tags: #<Tag:0x00007fd548512d70> #<Tag:0x00007fd548512ac8> #<Tag:0x00007fd548512690>


Hi there!

I use ImageJ and Fiji since a long time, but now I really need to keep safe the exif metadata in each image after processing --unless they become pretty useless in third party software i.e. Agisoft Photoscan. Up to now I’ve been using different tips and workaround e.g. exporting the exif as .csv or .txt file in the same directory, but that was for small dataset.

Let’s be honest: the problem of ImageJ deleting the image metadata is quite ancient. Unfortunately I haven’t got the appropriate skills and knowledge, so I ask you lovely community of nerdy devs if it’s possible to work on this feature and fix it.

I thank you in advance for any help!



Hi @babllz

can you state in detail, which meta data entry is lost and how this influences further processing? I just had a quick look in the list of exif meta data entries and now I’m wondering, what;s going wrong.

To justify a bit the way imagej is handling meta data: If in the meta header of a file is saved, that a certain light source was used to image this file (exif tag 0x9208), it is reasonable not to copy this meta data entry to a processing result image. Otherwise, other software would misinterpret the meta header. The processing result image was created using ImageJ and definitely not using this particular light source :wink:



Hi Robert,

thank you for your reply!

Well, I’ve just made a simple operation to show you an example.
I have

  • opened in Fiji a TIFF image (obtained from a RAW developed with Adobe Lightroom),
  • exported the Exif in CSV format (in Fiji File > Import > Exif data…),
  • saved the image in TIFF format without performing any processing,
  • exported the Exif again.

Here I attach the screenshots of the metadata before and after saving the image.

As you can see, ImageJ retains only some metadata on the image resolution, bit depth, and few other information. Third party software for photogrammetry (e.g. Photoscan) and image stitching (e.g. PTGui) strongly rely on metadata like focal length, lens and camera model etc. Without the metadata embedded in the images, these information must be added manually, with consequent loss of accuracy.

In any case, the value of an image – that is not computer-generated – relies also on the completeness of the technical metadata, and this loss is, in my opinion, unacceptable in a scientific context.


Hi @babllz,

Again, I disagree on this point. If an image is (potentially) manipulated within imagej and the meta header claims it was acquired with a camera, the header would be lying about the source of the image. I can imagine special situations where it should be possible to keep certain parts of the meta header, but copying everything should not be done by default.

To bring you closer to a solution of your issue: Did you try saving the important meta header entries e.g. using exiftool to a csv file and copying them using exiftool to the resulting tif file?



Nope, not yet, I’ll surely try this pathway asap!
I understand the technical point of view you highlighted, and imagine that we’re all destined to have images without technical metadata when processed with ImageJ ¯_(ツ)_/¯

Thank you for your reply!