I've become very interested in the translation of sound into image and image into sound. Recently I've been using a programme called "Bitmaps And Waves" or BW.exe coded by Victor "X" Khashchanskiy (http://www.elisanet.fi/victorx/) which applies a Fourier Transformation (http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/OWENS/LECT4/node2.html) to translate from the pixels in the image to the sound samples in the wav. The virtue of this method is that it's a true mathematical relationship and so it works both ways. That is to say, when the reverse process is applied to the image produced from the sound it reconstructs the exact same sound.
When some change is made to the image, such as a resize or a photoshop effect, turning that image back into the sound carries with it the mathematical distortion and thus changes the sound.
As an experiment I converted all of the tracks from the Studio Typewriter "Kitchen Sink" album into wavs and then used BW.exe to create bitmap images of them. Then I resized all 14 images to a uniform size (640x480 pixels) and then turned them back into wavs. Finally I converted the wavs to mp3s and uploaded them to last.fm. (As I write this I'm still waiting for some of them to upload - a last.fm glitch is blocking some of the uploads).
Here are the images created from the tracks: