In order to present the data in more visual way, each graph element can be represented by a unique image.
Due to the relativly large number of possible elements (i.e ~ 500K authors in DBLP, 1.5 Millon actors in IMDB)
images are not saved in the DB, but are found each time a new element is added to the graph.
Image search is done in server side, and actual image download is done in Client.
Image URL is passed in an XML from Server to Client.
Server side can be set to automatically retrieve matching images for each element (by the element name),
from one of two possible sources:
In DBLP implementation, images are searched only within the author's homepage or faculty domain,
in order to deliver the most accurate results.
In IMDB implementation, search is not limited to a specific site.
Google's returned results are in JSON format, and are parsed using PHP:JSON function.
In order to increase speed and reduce traffic bandwidth,
images are taken directly from Google's thumbnail servers (tbn3.google.com).