4 ms·

I don't like it. There are three purposes mixed in this "watermark": a) Give the innocent viewer a way to know the name of the photograph and go the original

by gbog 14y ago

I don't like it. There are three purposes mixed in this "watermark": a) Give the innocent viewer a way to know the name of the photograph and go the original source of the image, maybe to buy the picture. b) Make it harder for less-than-innocent people who want to take these images and use them for illustrative purposes without people noticing they did not actually take these pictures. c) Inject a fingerprint in these images so that can serve as a proof of ownership in case of legal issue, and that may also help crawling the web searching for them. For a), it seems much preferable to have a classic caption, where more useful information may fit (date, place, etc) For b), with my limited gimp skills, I guess I would need 30 minutes to remove the watermark the first time, and then it would take 2 minutes. For c) I suppose a real invisible watermark is actually much harder to erase and a better proof of ownership, but maybe they do add such a fingerprint also? Anyway, I see here yet another contradiction of DRMs: content provider want more and more people accessing their content for free, but they also want, at the same time, this access to be unpleasant enough so that these users would buy the paid content. I don't think this scheme will survive in the long-term.

shortformblog 14y ago
You realize that Getty Images used to have much uglier watermarks, right? Example: http://dl.dropbox.com/u/58607934/stuff/getty.jpeg http://dl.dropbox.com/u/58607934/stuff/getty.jpeg So in other words, option b and option c has always been on the table. However, we're in an era where images are becoming much easier to share on other platforms. All someone has to do to share a picture these days is to pin it or post it on Tumblr. All that metadata gets stripped when people do that. At least with this, people know where to find it. Considering the changes in sharing images, why wouldn't Getty want to take advantage of this? That way, if people are going to share them anyway, at least it isn't a total loss to them — and hey, people might even be more likely to share them because it doesn't have a supremely ugly watermark. Media outlets pay $100 — each — for many of these photos. This a big business for Getty, so they need to protect it. Now, Getty could do things to make it easier for smaller entities — say, bloggers — to take advantage of this. This feels like a good, if imperfect, compromise. Now you might be willing to use GIMP to modify a photo, but most people won't. It's why Adobe can charge $600 for Photoshop. Too expensive? You're not their target market. This isn't to say there's not a better way, but it just feels like we're wringing our hands over what seems to be a good-faith effort to improve something which previously hindered the experience.
- gbog 14y ago
  You say "much uglier" but this is a subjective judgment, right? In fact if I had the choice I may prefer the previous one, they may have a less annoying influence on the original photo.
  - shortformblog 14y ago
    I feel like you're just being contrarian at this point — I mean, you weren't even aware that this watermark existed until I pointed it out.
weinzierl 14y ago
Just as a remark to your point c: A watermark (visible or invisible) provides neither proof of ownership nor does it help crawling the web searching for the watermarked artifact. Anyone can add a watermark to an image, regardless if they are the owner, creator, copyright holder or whatever. Anyone can even add a getty watermark to an image. This is called a copy attack and it's the reason why the watermark is not very useful for identification by a crawler. A getty watermarked image could be a real getty image or any image someone inserted a getty watermark. In other words you could sort of DoS the getty crawler very easily. Digimarc had an interesting research paper on their website about this topic but unfortunately it had been removed. In 2011 I wrote an email to Digimarc customer support asking if their newest watermarks are still vulnerable to a copy attack and the answer was: "There are hundreds of applications and specific use cases for digital watermarking as I'm sure you found in your research. Some of our published papers describe theories and some describe applications and solutions that are available for license today as end-user products. " EDIT: Found the Digimarc paper http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CFAQFjAB&url=http%3A%2F%2Fmcrouter.digimarc.com%2Fdocs%2Ftechnology-resources%2Fdmrc_image_signatures.pdf&ei=EA_BT8SEE6mm4gS51qizCQ&usg=AFQjCNELe50YKlpVL0TeJHOtlIO64IqQWA&sig2=Nvmc3ibcacNQvH3Ju5t7jA http://www.google.com/url?sa=t&rct=j&q=&esrc=s&#....
- jonhendry 14y ago
  A Getty crawler would have access to all the images that really are Getty images, making it fairly easy to identify a non-Getty images with fake Getty watermarks. Some sort of hashing function, perhaps, or something like tineye uses. Since these new watermarks include a URL that identifies the Getty stock number of the photo, someone trying to fake that would end up putting a bogus stock number on the watermark, which makes it even easier to ignore fakes. The fake watermark would either use a number that doesn't correspond to a Getty picture at all, or a number that points at a different image.
  - weinzierl 14y ago
    I agree with you that the Getty crawler has access to all the Getty images and can identify non-Getty images. My point is that the watermark is not of much help for this task. On the one hand it is easily misled to find false positives on the other hand it will miss many images which have been altered strongly enough to destroy the watermark. You are right with your objection that watermark contains the stock number. I don't see however how this can help a crawler to identify a fake stock number. To compare the (possible altered) image with the original we have to use an image hash and then we could have used the image hash in the first place.

by gbog 14y ago

shortformblog 14y ago

You realize that Getty Images used to have much uglier watermarks, right? Example: http://dl.dropbox.com/u/58607934/stuff/getty.jpeg http://dl.dropbox.com/u/58607934/stuff/getty.jpeg So in other words, option b and option c has always been on the table. However, we're in an era where images are becoming much easier to share on other platforms. All someone has to do to share a picture these days is to pin it or post it on Tumblr. All that metadata gets stripped when people do that. At least with this, people know where to find it. Considering the changes in sharing images, why wouldn't Getty want to take advantage of this? That way, if people are going to share them anyway, at least it isn't a total loss to them — and hey, people might even be more likely to share them because it doesn't have a supremely ugly watermark. Media outlets pay $100 — each — for many of these photos. This a big business for Getty, so they need to protect it. Now, Getty could do things to make it easier for smaller entities — say, bloggers — to take advantage of this. This feels like a good, if imperfect, compromise. Now you might be willing to use GIMP to modify a photo, but most people won't. It's why Adobe can charge $600 for Photoshop. Too expensive? You're not their target market. This isn't to say there's not a better way, but it just feels like we're wringing our hands over what seems to be a good-faith effort to improve something which previously hindered the experience.

gbog 14y ago
You say "much uglier" but this is a subjective judgment, right? In fact if I had the choice I may prefer the previous one, they may have a less annoying influence on the original photo.
- shortformblog 14y ago
  I feel like you're just being contrarian at this point — I mean, you weren't even aware that this watermark existed until I pointed it out.

weinzierl 14y ago

Just as a remark to your point c: A watermark (visible or invisible) provides neither proof of ownership nor does it help crawling the web searching for the watermarked artifact. Anyone can add a watermark to an image, regardless if they are the owner, creator, copyright holder or whatever. Anyone can even add a getty watermark to an image. This is called a copy attack and it's the reason why the watermark is not very useful for identification by a crawler. A getty watermarked image could be a real getty image or any image someone inserted a getty watermark. In other words you could sort of DoS the getty crawler very easily. Digimarc had an interesting research paper on their website about this topic but unfortunately it had been removed. In 2011 I wrote an email to Digimarc customer support asking if their newest watermarks are still vulnerable to a copy attack and the answer was: "There are hundreds of applications and specific use cases for digital watermarking as I'm sure you found in your research. Some of our published papers describe theories and some describe applications and solutions that are available for license today as end-user products. " EDIT: Found the Digimarc paper http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CFAQFjAB&url=http%3A%2F%2Fmcrouter.digimarc.com%2Fdocs%2Ftechnology-resources%2Fdmrc_image_signatures.pdf&ei=EA_BT8SEE6mm4gS51qizCQ&usg=AFQjCNELe50YKlpVL0TeJHOtlIO64IqQWA&sig2=Nvmc3ibcacNQvH3Ju5t7jA http://www.google.com/url?sa=t&rct=j&q=&esrc=s&#....

jonhendry 14y ago
A Getty crawler would have access to all the images that really are Getty images, making it fairly easy to identify a non-Getty images with fake Getty watermarks. Some sort of hashing function, perhaps, or something like tineye uses. Since these new watermarks include a URL that identifies the Getty stock number of the photo, someone trying to fake that would end up putting a bogus stock number on the watermark, which makes it even easier to ignore fakes. The fake watermark would either use a number that doesn't correspond to a Getty picture at all, or a number that points at a different image.
- weinzierl 14y ago
  I agree with you that the Getty crawler has access to all the Getty images and can identify non-Getty images. My point is that the watermark is not of much help for this task. On the one hand it is easily misled to find false positives on the other hand it will miss many images which have been altered strongly enough to destroy the watermark. You are right with your objection that watermark contains the stock number. I don't see however how this can help a crawler to identify a fake stock number. To compare the (possible altered) image with the original we have to use an image hash and then we could have used the image hash in the first place.