Hacker News Reader

Comment by sethbannon | Hacker News Reader

From the article: "On Slate Jonah Peretti goes on to say ”would love if every image contained some secret metadata and a way to license that image. But the practical reality is that it is pretty challenging“."

Now there's an interesting startup idea.

View on HN

Digimarc did this years ago but the problem has always been enforcement, not detection. As noted in the article, BuzzFeed knew where the image came from and could have figured out a way to license it pretty easily- but they didn't want to.

http://en.wikipedia.org/wiki/Digimarc

Haha. What he meant was:

"I'd love it if it remained 'pretty challenging' to identify who owned what on the web, because our whole business model would crater if we had to pay for the content we rip off."

Exploiting other people's content is fundamentally baked into the business models of these guys and their "I-can't-believe-it's-not-plagiarism" cousins.

It already exists: http://en.wikipedia.org/wiki/Steganography#Digital

Unfortunately, it's easy to thwart. (just transform the image slightly)

Not at all. Enough methods, such as spread spectrum watermarking, survive a multitude of digital deformations and 'analog' attacks like printing and physically deforming. Action from 1.30 and German, but the demo is pretty clear. [1]

[1] http://www.3sat.de/mediathek/index.php?display=1&mode=play&o...

I can't watch the video, can you please provide some information/source about the technique.

I saw i paper a long time ago of a stenography quite resistant to transformations so the basic information was still extractable from the picture. Still i think that a photograph of the monitor would remove it.

EDIT: Even a photo filter might not succeed if it did a homogenous effect across the pixels.

I have to find that algorithm someday.

Knowing BuzzFeed, they would probably just apply a photo filter and call it a day.

That doesn't make it any easier to license, does it?

There's an industry effort to set up a universal licensing language and registry:

http://www.useplus.com/aboutplus/system.asp

https://www.plusregistry.org/

Though I can't say it's gained much traction.

Every photo on Flickr has a copyright tag attached to it. At least on Flickr it's not challenging at all.

This already exists:

http://www.photometadata.org/meta-resources-field-guide-to-m...

The problem is that the people you most want to follow this are the least likely to check IPTC metadata first.

You're missing the point. IDing the copyright owner is the easy part. Licensing the picture is the hard part - you have to write an email to the person and there's all this tedious back-and-forth whereas what the buyer wants is a one-click button like on Amazon.

Although the page you link to is a great resource, it's a sad fact that there's no metadata for 'Cost of 3-month nonexclusive license: $60 per million impressions. Remit-to: my_account@paymentprocessor.com'.

All sorts of restrictions on usage and what-all else, but no fields for price or payment method...the only two things that a potential buyer actually needs to transact business.

/facepalm

So in your scheme it would make sense to add some custom fields listing price and payment method rather than a URL where one could learn the current prices? In your hurry to make a teduously pedantic purported correction you might perhaps have misunderstood the problem: there are many ways to sell photos quite easily online but no way to force people to use one.

I disagree that that is the problem. There are many ways to sell photos online, but a photo editor doesn't want to be restricted to only browsing a pool such as Getty or Shutterstock (which are not open to all comers either). My argument is that while there are widely accepted mechanisms for establishing copyright status (eg CC-by-SA or sharealike) there's no universal standard to automate or streamline licensing, which is why (rent-seeking) brokerages persist.It's a shortcoming of the Creative commons approach that it doesn't include any sort of transaction mechanism for the CC-BY version. I went to Dan Catt's Flickr page; there's a statement of copyright but there's no link to buy a license for the photo. This is the problem.

Licensing is a pain in the ass because it's not automated; it's not automated because many artists are scared of pricing (and I say this as an artist). They're reluctant to stick a price tag on their work out of the fear that they'll pick one that's too low or too high and as a result there's no standardized way to license a an interesting piece of art you stumble across, other than by contacting the apparent copyright holder and trying to enter into negotiations, which is so time-consuming that it's often not economically efficient for the would-be licensee. It's cheaper to assume the business risk of occasionally infringing than to license IP without any standardized framework for doing so - especially for articles of this kind that are thrown together in the space of an hour or two. A URL is not helpful because (as with this example) we have no idea whether it will provide a mechanism to perform a license transaction or not.

What I'm proposing is adding term metadata that allow the information about pricing to travel with the picture like any other piece of metadata. It's kind of ridiculous to have all sorts of metadata codifying the creator/copyright owners ownership interest in the IP but not to include any on what that interest is valued at. If such term metadata were to become standardized in similar fashion to the Creative Commons licensing, then artists would be able to sell more of their work with less transaction overhead, while licensees would be able to buy with the same convenience that consumers enjoy.

> What I'm proposing is adding term metadata that allow the information about pricing to travel with the picture like any other piece of metadata.

This is a fundamentally naive view of the problem: the rates will vary over time and based on the usage and photographers will charge more for the front-page of a major magazine than their kid's PTA newsletter.

The solution is, again, a URL which allows all of those details be conveyed without trying to cram a soon-to-be-stale copy into the photo metadata.

Type:

    jhead -purejpg in.jpg out.jpg

And no pesky EXIF or IPTC to worry about. Hence digimark and others that put the info in the image.

That was the point of my last sentence: metadata exists for honest people who choose to do the right thing. Anyone dishonest can trivially strip this but the snake-oil vendors like digimark only raise that bar enough to catch the most inept thieves.

Put another way: Buzzfeed could trivially put a system into place to scan images to prevent mistakes. It would be a much harder problem to reliably stop dishonest editors – at that point, it'd be better to simply use something like TinEye or Google Image Search.

Until I take a screenshot of the image and then use that. Printers create a light band of yellow on every print that helps identify the printer it came from (counterfeiting with a store bought printer is stupid), maybe something along those lines would work?

It would be just as doomed as any other DRM scheme. You can't stop anyone from copying an image, removing any metadata that might be there, or even if they're clever, from cloning out a watermark.

Reading the original quote, and assuming good faith, they're wishing for some system that would permit them to establish an image's owner.

If you don't take steps to strip out typical metadata, such is indeed already present, in EXIF and IPTC fields.

IPTC http://www.iptc.org/cms/site/index.html?channel=CH0089

EXIF

You can just image search your photos every X months.

I agree, it could be a startup.

http://www.tineye.com/

Google images is better.

Reply on news.ycombinator.com