DSLR Photography Forum

Hey guys, just thought I'd share a little trick on finding out whether an image has been altered or not. These days you can never trust anything you see on the internet...

Take for example, this image here:


It looks like the photographer has captured the exact moment that the cat has its legs tucked underneath it, so that it looks like it has no legs / and is hovering in midair.

Grab the image URL:
http://i.imgur.com/2KwfT.jpg

And head over to Tineye, which is a reverse image search engine. Basically it finds occurrences of images on the web that match the input image.

Search result:
http://www.tineye.com/search/631b13cb918...c74ea4c716

Immediately you see a bunch of results, all of which show the cat with legs! Voila... fake indeed.



But go one step further, and click the "Compare images" link on one of the results. It'll bring up a little popup window, with a toggle button so you can flick between the original and the source. 'Shop exposed!


You can also use tineye to see if any of your images have been used anywhere else on the web, possibly without your approval.

I like more the original picture than the post processed.

Thanks for the link. I'll come back later to read a bit more about tineye.

I agree with Irma. Furthermore, there are tell-tell signs of post-processing in the image. leg removal was not done all that carefully. More to the point, I think that this search engine is great if you wish to find if anybody is stealing your photos. Unfortunately, the tool was unable to find my own most popular 3 photos on Flickr. I just wonder how well it would do with more obscure images on more obscure sites.

P

Another good way of telling if a picture has been PhotoShopped is to look for the telltale "Will" mark in one of the corners...

Toad Wrote:Another good way of telling if a picture has been PhotoShopped is to look for the telltale "Will" mark in one of the corners...

LOL.

It was obvious to me it had been shopped, when cats leap, they extend their front legs to "reach" for the landing, they don't took them back.

But ultimately better fun removing the cat's legs before the shot....take more than the patch tool in pp then, I'll warrant.

:|

Zig Wrote:But ultimately better fun removing the cat's legs before the shot....take more than the patch tool in pp then, I'll warrant.

:|

Probably a little messy and you could get scratched....

Came across this site as well which does Error Level Analysis:
http://errorlevelanalysis.com/

Quote:“Error level analysis (ELA) works by intentionally resaving the image at a known error rate, such as 95%, and then computing the difference between the images. If there is virtually no change, then the cell has reached its local minima for error at that quality level. However, if there is a large amount of change, then the pixels are not at their local minima and are effectively original.”

I tried it on the image above, but couldn't really see anything remarkable... oh well I guess it doesn't work 100% of the time.
http://errorlevelanalysis.com/permalink/7cbc515/

Jules - I think that's brilliant - I can totally see how that works. I'll bet that we can see it with our naked eye as well - but our brain doesn't have the capacity to do the analysis...

..and the cat photo actually has a quite distinct dark area where the shopping occurred - total give away if you believe the analysis of error patterns that they are proposing in the article. I'd love to put one of my multi image composites through that and see what it looks like.

There's one little problem with that method... I do all my work in PSD files, only converting to jpg when posting to the web.... so their is no degredation until the final save and all areas of the image have been saved an equal number of times.

EnglishBob Wrote:There's one little problem with that method... I do all my work in PSD files, only converting to jpg when posting to the web.... so their is no degredation until the final save and all areas of the image have been saved an equal number of times.

I would agree with you on that one - but I don't think that is their algorithm exactly.They are a bit vague about it - but I think that is analyzing areas for modifications such as brightening, cloning, and smoothing as well. I'd like to know more.

Interesting. Presumably then, it only works by extrapolating deviations within an image? In other words, if one had applied an effect that was "universal" in nature(ie, not like smart sharpen, smart blur, etc, it would appear "passed" as genuine? If I've understood this correctly(and I might not have), this would only flag up areas of presumed local deviation within the image...so that any lines of uniform deviation would appear and set alarm bells ringing. A darn handy tool, maybe, for news editors and competition organisers, and those who are unable to immediately spot fakes/shopped stuff(as I gather experts in image analysis develop an extraordinary eye in spotting them.

Zig Wrote:Interesting. Presumably then, it only works by extrapolating deviations within an image? In other words, if one had applied an effect that was "universal" in nature(ie, not like smart sharpen, smart blur, etc, it would appear "passed" as genuine? If I've understood this correctly(and I might not have), this would only flag up areas of presumed local deviation within the image...so that any lines of uniform deviation would appear and set alarm bells ringing.

That is how I interpret it. Your point about universal shopping would be totally valid if this is the case...