1

I want to remove exif from an image before uploading to s3. I found a similar question (here), but it saves as a new file (I don't want it). Then I found an another way (here), then I tried to implemented it, everything was ok when I tested it. But after I deployed to prod, some users reported they occasionally got a problem while uploading images with a size of 1 MB and above, so they must try it several times.

So, I just want to make sure is my code correct?, or maybe there is something I can improve.

from PIL import Image

# I got body from http Request
img = Image.open(body)

img_format = img.format

# Save it in-memory to remove EXIF
temp = io.BytesIO()
img.save(temp, format=img_format)
body = io.BytesIO(temp.getvalue())

# Upload to s3
s3_client.upload_fileobj(body, BUCKET_NAME, file_key)

*I'm still finding out if this issue is caused by other things.

Rza
  • 185
  • 1
  • 10

1 Answers1

3

You should be able to copy the pixel data and palette (if any) from an existing image to a new stripped image like this:

from PIL import Image

# Load existing image
existing = Image.open(...)

# Create new empty image, same size and mode
stripped = Image.new(existing.mode, existing.size)

# Copy pixels, but not metadata, across
stripped.putdata(existing.getdata())

# Copy palette across, if any
if 'P' in existing.mode: stripped.putpalette(existing.getpalette())

Note that this will strip ALL metadata from your image... EXIF, comments, IPTC, 8BIM, ICC colour profiles, dpi, copyright, whether it is progressive, whether it is animated.

Note also that it will write JPEG images with PIL's default quality of 75 when you save it, which may or may not be the same as your original image had - i.e. the size may change.

If the above stripping is excessive, you could just strip the EXIF like this:

from PIL import Image

im = Image.open(...)

# Strip just EXIF data
if 'exif' in im.info: del im.info['exif']

When saving, you could test if JPEG, and propagate the existing quality forward with:

im.save(..., quality='keep')

Note: If you want to verify what metadata is in any given image, before and after stripping, you can use exiftool or ImageMagick on macOS, Linux and Windows, as follows:

exiftool SOMEIMAGE.JPG

magick identify -verbose SOMEIMAGE.JPG
Mark Setchell
  • 191,897
  • 31
  • 273
  • 432
  • How about saving data in-memory and replace the body before upload to s3 e.g. `body = io.BytesIO(temp.getvalue())`, is it fine? – Rza May 16 '22 at 07:46
  • Sorry, I don't know what you mean. Saving an image in-memory doesn't discard any EXIF data any more than saving an image to disk. – Mark Setchell May 16 '22 at 07:52
  • Because I need to upload data to s3 `e.g. s3_client.upload_fileobj(body, ...)`, so I have save it first right? – Rza May 16 '22 at 07:57
  • So, the steps are 1. remove exif data from image, 2. save in-memory, 3. upload to s3 – Rza May 16 '22 at 07:59
  • My code will strip the EXIF data and leave you with a stripped `PIL Image`. You can then use your three lines of code beginning `temp = io.BytesIO()` to save it and send it to S3, yes. – Mark Setchell May 16 '22 at 08:44