Skip to content
ImageDoctor
Guide6 min read

Watermarks that are hard to remove

No visible watermark is impossible to remove. The aim is to make removing it cost more than the picture is worth to whoever wants to reuse it — and that depends on how much of the image the mark covers, what is underneath it, and whether it is identical on every photo.

Processed locally in your browser

A small logo in the corner credits your work but is the easiest kind of mark to get rid of. A mark that is hard to remove covers the picture, sits over detail rather than empty space, and is large enough that cutting it out would ruin the image. The Watermark tool can do all three; this guide explains why they matter and where a visible watermark stops helping.

How visible watermarks get removed

  • Cropping. A mark in a corner or along one edge goes with that edge. Most photos can lose a strip at the side without anyone noticing.
  • Painting it out. Photo editors can fill a selected area using the pixels around it. This works best when the mark is small and sits over something plain — sky, a wall, a studio backdrop — because there is little detail to rebuild.
  • Working it out from many copies. Researchers at Google showed that when the same watermark is applied the same way to a whole collection of images, software can estimate the mark and its transparency from the collection, then remove it from every image automatically. In their tests, moving the mark to a random position on each image, or making small random changes to its opacity, did not stop this. Slightly distorting the mark differently on each image did make it much harder (Google Research, 2017).

What makes a watermark harder to remove

  • Cover the whole picture. Switch on Tile across the image and the mark repeats over the entire surface, rotated as you set it. There is no clean edge to crop to; removing it means repairing the whole photo.
  • Put it over detail. A single mark across the subject — the Center position, or a middle one — is far more work to paint out than one over empty sky in a corner.
  • Make it big enough to matter. Text Size is a percentage of the image's shorter edge, and logo Width a percentage of its width. A mark that spans a good part of the picture can't be cropped away without losing the picture.
  • Keep it legible on any background. An Outline or Drop shadow keeps light text readable on bright areas, so you don't have to move the mark to a plain patch where it would be easier to remove.

There is a trade-off with Opacity. A fainter mark spoils the photo less for the people you want to see it, and is also easier to hide. Preview at full size and choose the strongest mark you're willing to show.

Starting points

You wantSettings to start from
A credit on photos you shareText or logo in a bottom corner, at the defaults (text 6% of the shorter edge, 60% opacity, 3% margin)
Proofs for a client before they payTile across the image, rotated, at a low opacity, so the whole photo is visibly marked but still easy to judge
One image you particularly want to protectA large mark over the subject, at Center, with an outline

Because sizes are percentages, the same settings look the same on a small phone photo and a full-size camera file, which makes it safe to watermark a whole folder at once in the Watermark tool or as one step in Batch Image Tools.

What this watermark does — and doesn't

  • It is part of the pixels. The mark is drawn into the image when it is saved; there is no layer anyone can switch off. Your original file isn't changed: you download a new copy, named like photo-watermarked.jpg.
  • It is identical on every image in a batch. That consistency is convenient, and it is also exactly what the collection attack above relies on. ImageDoctor doesn't randomise or distort the mark per image, so for a large public archive treat a visible watermark as a credit and a deterrent, not a lock.
  • It doesn't remove the photo's details. When a JPEG is watermarked and saved as JPEG, its EXIF — camera, date and any GPS location — is carried over to the new file (the embedded preview thumbnail is removed). Saving as another format leaves it behind. To share without that information, run the watermarked photos through the Metadata Remover.

Beyond a visible mark

A watermark tells people whose photo it is; it can't stop anyone copying what is on their screen, and nothing that runs in a browser can. What limits reuse most is not publishing more than you need to: post a copy at the size it is shown, using the Image Resizer, and keep the full-resolution original to yourself.

Common questions

Yes. A corner mark can be cropped off, a small mark over a plain area can be painted out in a photo editor, and a mark applied identically across many images can be estimated and removed by software. A large or tiled mark over detailed parts of the picture takes much more work.

For protection, yes: a tiled mark covers the whole image, so there is nothing to crop to. A single corner mark is better when you only want to credit the photo and keep it clean to look at.

In ImageDoctor, a JPEG saved as JPEG keeps its EXIF, including any GPS location, while other output formats drop it. Run the result through the Metadata Remover if you don't want those details shared.

Tools in this article

All guides