Page MenuHomePhabricator

Resized raster image does not succeed metadata of the original file
Closed, DuplicatePublic

Description

When the image file is resized or rendered to a raster image (from vector source) by wiki software, the metadata generally does not get carried over even if the original file contains any. While I understand there is some major argument of how metadata should be added on Wikimedia platform, I think we at least should transfer the original metadata to the resized raster render instead of leaving the metadata fields totally empty.

After the change of Wikimedia that the file description page no longer provides original size preview if the width of the original raster image is too long, the down-scaled version is always the one being shown, which means readers are more likely to have downloaded the metadata-less version because the "original file" link isn't that obvious to the user that it is actually different than the preview image in many ways.


Version: unspecified
Severity: enhancement

Details

Reference
bz65475

Event Timeline

bzimport raised the priority of this task from to Low.Nov 22 2014, 3:22 AM
bzimport set Reference to bz65475.
bzimport added a subscriber: Unknown Object (MLST).

Thanks for taking the time to report this!

What would be steps to follow if I wanted to reproduce? Any testcase?

(In reply to Andre Klapper from comment #1)

Thanks for taking the time to report this!

What would be steps to follow if I wanted to reproduce? Any testcase?

Steps to reproduce:

  1. Access the description page of any oversized PNG file with metadata at Commons like File:FutureMTRNetworkAfterMerger.png
  2. Download both the original file and the resized preview.
  3. Use any application capable of reading Exif/XMP like ExifTool to read both version of the image.
  4. Compare the metadata between the 2 files
  5. The resized file contains fewer metadata than the original file.

Expected: resized image should contain all metadata from the original file.

I dont think thats the expected result. Lots of useless crap in the metadata.

However there are some fields which would be nice to copy over like artist

We also add a field to thumbnailed images to link to the source (but possibly not with vips, so not for really big pngs. Probably also not for svgs).

Im confused about if you are complaining about svgs (vector) images specificly or all files. You talk about vector formats in comment 0 but pngs (raster) in comment 2.

After the change of Wikimedia that the file description page no longer
provides original size preview if the width of the original raster image is
too long, the down-scaled version is always the one being shown.

That's incorrect, the situation is not new. This is the way its always been. We show the original image asset where possible, but we're not display huge files inline on the image description page. Some of those original image files are hundreds of megabytes big. There's even a tiff file which is 1 GB big.

There is already the discussion 17503 about generating metadata by checking licensing/author info or information template from the media description page, but the progress is disappointingly slow. Rubbish or not, it's still better than containing no metadata at all in the resized file. Please prove that metadata is ever the culprit of oversized media file.