I think any discussion of "transcoding" image (or other data) into sound has to take into account that such mapping is purely subjective. The artistry is in finding meaningful transpositions from one medium to another. Using "raw" data as sample values is certainly one way, and using an image as spectrographic values for resynthesis is certainly another. There are certainly enough softwares out there that do either trick, and those sounds are familiar to us from the many many artists who have used those softs already (to death in some cases). Dig a little deeper and decide for yourself the relationship of pixel to sample instead of relying on other people's ready-mades and you might be on to something new.

Anyway, I'll be giving a workshop with Sara Kolster on PD and GEM in Bergen Norway next month, and one facet of that workshop is the relationship of (moving) image to sound. Emphasis will be placed on direct connections between sound and image through three different methods: influence of sound on image, influence of image on sound and the use of common "control data" for both image and sound. Full text has been posted on the Microsound-announce list:

http://or8.net/pipermail/microsound-announce/2005-April/000080.html

derek