This stems from a bit of concept coding for a design but it's been adapted and extended to now use a PNG as a source mask and some tidy up of the methods
Essentially it takes in a list of images and a source PNG file, creates a bitmask to the size of the image and then uses that to output html (using cherrypy) where the image content for each cell comes from the passed images.
You could for example feed it a flickr stream, the picture urls from a facebook graph call or a list of twitter profile pics. The current example uses the results of a flickr search to create a source table of images for use.