How to Tell if an Image is Edited
How can you tell if an image is photoshopped or edited? There are three categories of techniques we use. Metadata, Metadata Validity, and Pixel level. Our image edited tool
checks only basic metadata, but our advanced service will subject your image to a full barrage of tests. Contact us
for a price quote.
Image files typically have a header that describes the contents of the file, the camera or tool that created the file, where it was taken, and any comments or license the creator has applied to the image. Checking metadata requires the original file. You cannot scan a photo or take a screenshot and do any metadata checks. Our image check tool
does all the basic metadata checks.
Camera or Scanner model
The exif data will contain the name of the type of scanner or camera that created the image. Seeing this makes it less likely the image is edited, but some editing programs preserve this data, so it doesn’t mean the image hasn’t been edited.
Exif data can record when an image was edited. When this doesn’t match the creation date, it means the image was edited after it was created.
Each camera or scanner has particular color profiles. The metadata records the color profile the image is encoded with. Some websites change these color profiles. We can recognize these website color profiles and tell you when we find them.
Does the actual image match what the container describes? If you saw a book with the cover title claiming it was Moby Dick but when you open the book, it was full of cartoons, you’d know something’s wrong. In the same way, when the metadata container description of the image doesn’t match what the actual image is, we know the image has been changed.
If the color model of an image doesn’t match the actual colors used, it has been edited. Some programs encode colors in a particular way, and we can compare the color model with the camera or scanner the metadata claims created the file. If they don’t agree, the image has been edited. The color model can act as a fingerprint for certain programs or cameras that makes it easier to track down how the image was created.
Rarely, an image will be resized and the program does not update the exif data. This is a clear pointer to an edited image.
Some cameras offer advanced features to change images after they are taken but before they are downloaded. These in-camera edits can leave tracks in the metadata we can follow.
Pixel Level Tests
JPEG Error Level
When a jpeg format image is loaded and saved again, the old layer of data degrades slightly, while the new edits are fresh. Looking at the error levels the image contains can show layers of edits clearly. This error level analysis (ELA) can pinpoint changes that have been made to the image, so even if it was edited, you may be able to find which part was edited. If it was sharpened, or recolorized, these can show up in ELA.
Color Distribution Discontinuities
If an image is expanded, the color distributions separate like teeth on a comb. This technique can detect when images have been resized. In some cases, the color distribution can even show if a piece of an image has been pasted in at different resolutions.
If a photo is created by putting together several images, the color channels may not match exactly. The mismatch can pinpoint pieces of the image that were photoshopped or edited in some way.
We can tell if pieces of an image were cloned from other parts. This is a common trick photo editors use to make it look natural to the eye and to cover up other edits. Copied regions may be difficult for a human to pick out, but a computer can automatically scan for bands or regions of copied pixels that point to use of the Clone Brush or copying and pasting.
Many photographers edit their photos and increase sharpness of regions they want to highlight. Alternately, they blur regions they want viewers to ignore. The grain of how sharp areas are can show where they have edited in this way.
Light Direction Mismatch
Light behaves according to the laws of physics. Unless the photo has been edited. Badly photoshopped images can have traces of differing light sources that would not be possible in the real world. This is a much harder test, as tools for finding light gradients rely on scene interpretation.