For a clever variation on PageRank using vision-based page segmentation, see:

http://research.microsoft.com/research/pubs/view.aspx?tr_id=754

For a more sophisticated web page similarity measure inspired by PageRank, see:

http://www-cs-students.stanford.edu/~glenj/simrank.pdf

For a ton of very cool papers on these sorts of topics, check out:

http://dbpubs.stanford.edu:8090/aux/index-en.html