Cracking w/ photoshop

[Soda_Popinsky]

Every now and then you will see a photo with a blurred out image. I was wondering what technology exists that allows you to clarify an image and retrieve info?

Take this for example:
http://www.oreillynet.com/network/20..._5_results.jpg

I've been messing around with the blurs in photoshop trying to retrieve the text. I had a pretty good result but it got lost in history, if I can get it back I will post it.

Maybe I'll post a challenge too if anyone is interested?

[phishphreek]

It depends on how it was edited. If it was edited in m$ paint... good luck getting it back.

If it was edited in photoshop with layers, then you can simply remove the layer.

I've seen a lot of places put the "watermarks" on the photos with photoshop as another layer.
I simply can highlight the watermark and remove it.

I know practically nothing about imaging and stuff like that. I've just messed here and there.

[valhallen]

phishphreek80 that would only work if you had access to the .psd file. When you save a file as a .jpg photoshop flattens the image and merges the layers.

As for getting rid of the filter - good luck! lol
esp on text because you cant really look at nearby pixels and try and guess what it should be.

v_Ln

[Soda_Popinsky]

I put this in GCC because I hardly think this should be focused on for more than a few minutes or for entertainment purposes

Here are my ideas (sample image attached):

Both Photoshop and Fireworks have a blur filter. If the artist used the blur filter repeatedly, then we may be able to reproduce the same results. A special software may be able to look closely at the blur and determine the algorithm that was used in producing it, therefore identifying the software used (or by looking at the file itself, jpgs made by photoshop say adobe in a hex editor or strings). From there, (attachment), you reproduce the font and possibly compare letter to letter, pixel to pixel... Also, if you look at the some letters, they come up lighter (c, l, i,), or lower (p,g) and some letters are close together creating darker spots in the blur.

So by creating an alphabet, and reproducing the filter, software may be able to compare the letters with each other to make matches. This is why I would expect software like this might exist. Of course, a gaussian blur from Photoshop would be tougher to crack because it allows a range of blur, but that could possibly be identified by comparing the font used with the blur found.

Similar to comparing MD5 hashes with a brute force method, I would expect you can compare images distorted by filters if you know how they were produced.

Of course, multiple filters and you're outta luck (or select->fill, for that matter)

So thats my idea.

[xierox]

Interesting idea. But I think that anything over a minimum blur would make it completely unretrievable as the letters would disolve into little blobs with very, very little differences. (Too small, I believe, for a computer to accurately bring them back.) Just my thoughts.

- Xierox