This is only partly true. Auto focus, at least the optical variety, is not possible without the optical hardware to do it. However, it is (at least in theory) possible to remove some blur using multiple frames, especially if there is a little bit of motion between them. An app that would have this capability would need some very low-level control over the camera.
Originally Posted by theweaselslayer
Furthermore, image stabilization is totally possible via software, but this again requires multiple frames, and probably similar low-level control over the camera.
So if this new camera api provides really good control over the camera, there are some really interesting image processing tricks that could be played. I'd like to see HDR images too.
All of these require some fairly substantial number crunching, so access to the GPU will be important too.
Edit: By the way, low-level access should also enable a bar-code scanner that could operate in situations where individual frames don't have good focus.