An astronomy friend told me that there is software (no links, sorry) that combines several pictures, each slightly different, to produce a better picture than any of the inputs.
Then I was watching the Photosynth demo yesterday.
Then Boing Boing had this post about REM’s Dubiln concert and they were encouraging everyone to record the audio.
Is it possible to take several crappy audio recordings and get some software to process it into higher quality audio? This is kind of what the Very Large Array does but as I understand it they know the precise location of each receiver to calculate the differential between the data streams. I’m thinking that the concert input would be from various unknown locations (possibly moving) in the concert hall and would not have a time registration. Has this been done already?