In this all-digital world, it’s easier for someone to do photo editing. Ordinary photo becomes extra-ordinary. Black can be white, fat turns to thin, young looks older, and the opposite. Then how we can find whether someone’s photo is real or has been through editing process?
Today SOT wants to share a method to analyze whether a photo is real or fake using ELA (Error Level Analysis). To do it, we will use FotoForensics.
Upload Photo to FotoForensics
First of all, visit FotoForensics site here:
Then insert URL or upload photo you want to research its originality.
After it is uploaded, on your left you will find four menus: ELA, JPEG %, Metadata, and Original.
- ELA is Error Level Analysis. Its function is to find out level mistake level in an image. This ELA analysis is the one you can use to find out whether the picture you investigated is real or has been through editing process.
- JPEG % is an analysis used to find out about picture quality in the last time it is saved. JPEG quality is so important to be found to decide whether the image you investigate is worth to be tested its ELA level. The higher a picture quality, there would be more accurate and easier for you to research it using Error Level Analysis.
- Metadata is photo data analysis which commonly contained various details about camera used, editor software used, etc.
- Original display the photo you want to research
Photo quality test
By using JPEG %, make sure the photo quality you wanted to investigate is not too bad. Avoid to test photo with quality below 70%. If the photo’s quality you tested is too low, find the same photo from other source which has higher quality. You can use TinEye or Google Image to find similar photo but with higher quality.
Error Level Analysis
This is the most important part from this photo analysis process. Here you can estimate whether this photo has passed editing process or not. You can also find out which part has passed editing touch.
In ELA, all parts supposed to have the same error level. The same surface, same color, same pattern, and other similar parts, they are supposed to have the same error level.
By ELA principle, let’s analyze girl’s photo which has been uploaded above.
First, on the picture, its ELA background seemed totally black, while the person’s photo is not. It means the background quality has different level quality compared with the person’s photo. It can be concluded that the background has through digital editing process… It seems like the editor whitened the background so it seemed brighter.
Second, let’s see the red colored part. It can be seen in the girl’s hat, its ELA is brighter than the clothes behind her back. Besides, there’s also clothes’ rope which its ELA is far more brighter than other parts. It can be concluded that the clothes area has been through editing process too. Probably the clothes’ rope has been edited to be seen brighter. It also happened to the girl’s red hat. It is a possibility that the clothes red color and the hat actually darker than it’s seen in the photo.
Third, let’s we put focus on her face. It can be seen that area on eyelid has brighter ELA than in chin and lips. It is proven that eyelids area also passed editing process.
Fourth, you can also see the eyes’ area. It can be seen brighter ELA in the middle of her eyes. It seems like, the girls’ eyes actually staring into different direction with what shown on the final photo. The model’s eye also has been edited to be directed into different point of view.
Fifth, see on the left arm and face of the ELA. It can be seen there are so many blue dots and red dots. Those are rainbowing. This rainbowing can be seen often when a photo is edited using Adobe’s products like Photoshop, etc. for GIMP itself, it didn’t produce this rainbowing effect. It’s proven that the photo has been edited using Adobe software. This evidence is strengthen by Metadata info which mentioned that last time, this photo was stored using Adobe Photoshop with “Save for Web” (not Save As) menu with 60% quality from its original photo.
From the analysis process of Error level and metadata, it can be concluded that the photo has experienced editing by using Adobe Photoshop on hat, eyes, clothes’ rope, also background. The photo in the last time saved using Save-for-Web method with 60% quality lower than its original photo.
How about that? It’s not quite difficult to investigate photo quality, isn’t it?
By the same method, you can investigate photo originality started with your lover’s photo in Facebook, even celebrities photos. Even you can also find out, from these two photos below, whether the girl actually young and edited into old, or the opposite…:D