A few months ago I began playing with manually corrupting jpeg files to see what kinds of artefacts I could create. I selected an image, compressed it down to a very small size (so I could easily manipulate large chunks of the data), opened it in a text editor (I like SubEthaEdit and TextMate) and added random text, copied pasted and generally shuffled the data, occasionally saving as new files.
Above is a QuickTime movie which animates through 12 of the resulting jpeg files. Click it to stop if it’s giving you a headache 
I re-compressed the jpegs just to be sure they wouldn’t crash QuickTime Player. Manually introducing errors and noise into files and then playing them is one of those “make sure you save any important files you have open” situations as things can grind to a halt.
I was playing with these images at Plug N Play at Kent St on Thursday night. I mentioned that I was planning on writing a php script which would similarly screw with jpeg images online and Sean told me about glitchbrowser.com.
From the site:
Computers are not allowed to make mistakes.
The glitch browser represents a deliberate attempt to subvert the
usual course of conformity and signal perfection. Information packets
which are communicated with integrity are intentionally lost in
transit or otherwise misplaced and rearranged. The consequences of
such subversion are seen in the surprisingly beautiful readymade
visual glitches provoked by the glitch browser and displayed through
our forgiving and unsuspecting web browsers.This work was produced for New Langton Arts Packets programme,
by Dimitre Lima, Tony Scott and Iman Moradi.
Glitch Browser lets you enter a website’s URL and it will show you the page with all of the images randomly glitchified. For example, here’s the most interesting photos from Flickr through the Glitch Browser. Great stuff!
Have you seen this?:
http://abstrakt.vade.info/?p=48