Shedding light on visual verification
By Gerri Berendzen
How do you know that a photo is real?
That’s the questions Samaruddin Stewart, project lead at Verified Pixel and with the SPJ-Google News Lab, asked to start out the “Visual Verification” session at ACES 2016.
The answers varied from “I’d seen it before” to “it just didn’t look right.”
“Generally speaking, when you see pictures with sharks, they aren’t real,” Stewart said as a photo of subway sharks appeared on the screen.
But Stewart’s answer is that for a long time the answer really was that there wasn’t a good way to tell. But today many of the user-generated photos we see on social media and the internet can be run through a series of tests to tell if they are real.
The goal of his work is to make sure newsrooms know that free tools exist online to help verify images.
Image verification has three levels, he said. File tests look at JPG data and can be automated. Footprint tests look at the content. Flaw test look at things like lighting and geometry that may show a photo is altered. The level many newsrooms can use to verify photos is the File level.
Editors can start with image search like Tineye and Google Reverse Image search. Stewart recommends using both because they index photos in different ways. Tineye looks for the same image and will show you alterations. It also indexes photos in collections. Google Reverse Image looks for photos that are visually the same.
After using image searches, Stewart recommends looking for EXIF metadata. He suggested using Jeffrey’s Exif viewer as free online tool. It shows EXIF data (such as file size, type of camera, geocodes, etc.) and even shows the location in a Google map, so you can tell where the photo was really taken from. And it can be used with Google Earth Pro, which is also useful in determining if a photo was taken were it was purported to be taken.
“Most of these tools work best when they’re stacked on top of each other” to provide multiple types of tests for a photo, Stewart said.
They are free tools that help you make better informed decisions about the photos you see on social media or that are submitted to you, he said.
Another tool, izitru.com, looks at the file and runs tests to see if the photo has been altered. And, finally, Stewart said, it’s important to analyze the sender of the photo with tools like Pipl, Email Sherlock and other people searches.
Stewart’s Verified Pixel project will allow people to send photos by email to the project, which will analyze those photos.
But Stewart acknowledges that these tools can’t do everything. There can be a real photo that doesn’t pass the tests and it doesn’t address things like traditional cropping of photos — which has always been around for news photographers.
He said the goal is to apply some criteria to UGC content.
“I would hope that the same criteria used for professional news photographers is used for user-generated content,” Stewart said.
