They could also put ads inside street view: use image-to-text to detect billboards and replace the part of the image with ads from Adsense.

Alexander: Wow, awesome suggestion :D

Good idea, but would that be allowed?

The Street View pictures are awfully blurry to do any OCR on, I'd imagine. Besides, does Google Maps really need to be cluttered with "FREE WAX WITH NAILS" or "Try the new Vivanno" or "I'm lovin' it" all over the place? I think not.

Just for fun, thought I'd mention: Fubar, not foobar, is actually the self-styled "only western bar" in Urumqi, Xinjiang Province, China. Over-priced, but not bad. You'd have to have a pretty good font recognition system to grab that data though when Google gets to Xinjiang. See pic

http://picasaweb.google.com/florence.ungaro/UrumqiVille/photo#5121581375617952306

I tried it using tesseract (http://code.google.com/p/tesseract-ocr/) and it didn't work at all... The OCR engine expects small black on white characters, it seems..

Google Street View could be used in a lot of different scenarios. Imagine Google Street view Law Enforcement edition. It could use face recognition software and run it through all of the images with people in them. Google could do this with a separate version of street view images since they are now blurring faces on the main site.

It can remove people from the images altogether by merging pictures taken over time.

Not sure if there are any official sources that would be able to provide these. But google earth would be a great media for bicycal paths.

Google's patent application on using image to text in Google Street view provides a number of details on how they could use OCR:

Database assisted OCR for street scenes and other images
http://appft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&u=%2Fnetahtml%2FPTO%2Fsearch-adv.html&r=1&p=1&f=G&l=50&d=PG01&S1=20070140595.PGNR.&OS=dn/20070140595&RS=DN/20070140595

I do like the idea of providing translations in streetviews.