Tumblelog by Soup.io
Newer posts are loading.
You are at the newest post.
Click here to check if anything new just came in.

October 13 2010

16:00

From tornadoes to Noxema: Hany Farid on using digital forensics to assess the authenticity of photos

Late last month, when tornado-like, end-times-are-nigh-style winds sliced through New York City, Time magazine posted to its NewsFeed an image of a twister forming in the waters beyond the Statue of Liberty — menacing, dark, grainy. The story the mag published — “Gotham Tornado: Amazing Photo of Twister Passing Statue of Liberty” — let the image in question pretty much speak for itself, with the only text accompanying the photo being its headline and twelve SEO-friendly tags. Turns out, though, that the graininess of the image was a symptom not of camera-phone authenticity, but of old age: The photo was shot in 1976.

Time’s mistake — the overzealous posting of an image that, given its context, would seem to be authentic — is one that almost any news organization could make. How, after all, do we check the accuracy of news images, particularly in moments of breaking-news urgency? Though we’re seeing a blossoming of fact-checking in text-based journalism, we have yet to see an equivalent movement in image-based news reporting — mostly, of course, because we lack good tools for determining whether images are authentic or manipulated, whether they depict what they claim to or something else entirely. Which is disturbing, given that we look to images — the raw material of the world, supposedly filtered only through a camera’s lens — to give us an unvarnished view of human events.

Enter Hany Farid. A computer science professor at Dartmouth, Farid is a pioneer in the field of digital forensics, figuring out how to analyze images to determine their authenticity. (Think CSI: Photojournalism.) In his day-to-day work, Farid deals with the algorithmic aspects of photo manipulation — how to translate light and shadow, for example, into data sets that will detect whether a particular image could actually exist in the real world.

“The issues of photo manipulation are going well beyond just the technical, mathematical, geek stuff,” Farid notes. “We’re struggling as a society to deal with what happens when this thing that we have learned to trust over so many years” — the photographic image — “becomes incredibly untrustworthy. And that, to me, is extremely interesting.”

The biggest element of mistrust is the increasing prevalence of manipulation, via Photoshop and other tools. While, most often, those programs are used to create basic composites, or ironically derivative images (cf: Rebecca Gayheart’s Noxemas, courtesy of Gawker), more and more, they’re also being used to produce composites designed to mislead the viewer. Take a tabloid photo of Angelina Jolie and Brad Pitt on a beach, which seems, at first, true to its caption: “CAUGHT TOGETHER!” Study the image closely, though — actually analyze it — and it becomes apparent that the photo is doctored: Its stars are lit from opposite sides. The image is a combination of two separate photos of the couple merged together through the magic of image-layering. (See more about this photo here.)

One problem is that our brains simply don’t seem to be wired for the kind of bit-by-bit observation such analysis requires. On the contrary: When it comes to images, our minds — which tend to interpret images as singular units, rather than composites of atomic ones — can often abet manipulation. “Your brain is remarkable, but it has some pretty serious limitations,” Farid says. And one of those limitations is image assessment. On sabbatical last year in Berkeley, he worked with neuroscientists and vision scientists, studying the limits of the visual system. In general, “people tend to over-trust their eye…and that’s a very dangerous game to be playing.”

Now, photo-tampering is becoming so prevalent, Farid notes, that mistrust of images is slowly becoming our default. “There’s almost a backlash,” he says, “and now there’s this over-skepticism of everything out there. It’s amazing.”

Digital forensics — shifting the analysis of images from an art to a science — is one way for images to earn back our trust. Forensic techniques “can help bring some sanity to this.” And the field “just continues to get more and more sophisticated,” Farid notes. “We’re able to do things today that a few years ago seemed unimaginable. And a few years from now, we’ll do something that, today, seems unimaginable.” Farid and his team have been developing software that can be used — by news organizations, in particular — to analyze the authenticity of images before publishing them. It’ll be a few years before that software is ready to be used, he notes; but “we are just getting to the stage where, I think, commercialization is viable.”

That’s a good thing, because the need for rigorous analysis of images is, and will continue to be, increasingly urgent. “It used to be, if you had a handful of photojournalists around the country, you could have some kind of quality control,” Farid points out. But now — “when everybody’s got a cell phone with video and images and they’re posting on their blogs, and to Twitter and YouTube” — the ethics of photojournalism are being tested and shifted. An analytic platform could bring a sense of universality back to those principles. “The issue with photo manipulation is that it’s not black-and-white,” Farid notes. “There are certain types of photo manipulation which are completely acceptable — and there are other ones that are completely unacceptable.” Images have always held a powerful place in journalism, of course, and “you’ve always been able to editorialize with photographs.” But now, Farid says, “it’s a question of degree.” Now, using images, “you can really change the entire story. And that, obviously, is a very different beast.”

March 19 2010

09:25

Interview: Nicolas Kayser-Bril, head of datajournalism at Owni.fr

Past OJB contributor Nicolas Kayser-Bril is now in charge of datajournalism at Owni.fr, a recently launched news site that defines itself as an “open think-tank”.

“Acting as curators, selecting and presenting content taken deep in the immense and self-expanding vaults of the internet,” explains Nicolas, “the Owni team links to the best and does the rest.”

I asked Nicolas 2 simple questions on his work at Owni. Here are his responses:

What are you trying to do?

What we do is datajournalism. We want to use the whole power of online and computer technologies to bring journalism to a new height, to a whole new playing field. The definition remains vague because so little has been made until now, but we don’t want to limit ourselves to slideshows, online TV or even database journalism.

Take the video game industry, for instance. In the late 1970’s, a personal computer could be used to play Pong clones or text-based games. Since then, a number of genres have flourished, taking action games to 3D, building an ever-more intelligent AI for strategy games, etc. In the age of the social web, games were quick to use Facebook and even Twitter.

Take the news industry. In the late 1970’s, you could read news articles on your terminal. In the early 2010’s you can, well… read articles online! How innovative is that? (I’m not overlooking the innovations you’ll be quick to think of, but the fact remains that most online news content are articles.)

We want to enhance information with the power of computers and the web. Through software, databases, visualizations, social apps, games, whatever, we want to experiment with news in ways traditional and online media haven’t done yet.

What have you achieved?

We started to get serious about this in February, when I joined the mother company (22mars) full-time. In just a month, we have completed 2 projects

The first one, dubbed Photoshop Busters (see it here), gives users digital forensics tools to assess the authenticity of an image. It was made as a widget for one of our partners, LesInrocks.com.

More importantly, we made a Facebook app, Where do I vote? There, users can find their polling station and their friends’ for the upcoming regional election in France.

It might sound underwhelming, but it required finding and locating the addresses of more than 35,000 polling stations.

On top of convincing a reluctant administration to hand over their files, we set up a large crowdsourcing effort to convert the documents from badly scanned PDFs to computer-readable data. More than 7,000 addresses have been treated that way.

Dozens of other ideas are in the works. Within Owni.fr, we want to keep the ratio of developers/non-developers to 1, so as to be able to go from idea to product very quickly. I code most of my ideas myself, relying on the team for help, ideas and design.

In the coming months, we’ll expand our datajournalism activities to include another designer, a journalist and a statistician. Expect more cool stuff from Owni.fr.

Older posts are this way If this message doesn't go away, click anywhere on the page to continue loading posts.
Could not load more posts
Maybe Soup is currently being updated? I'll try again automatically in a few seconds...
Just a second, loading more posts...
You've reached the end.

Don't be the product, buy the product!

Schweinderl