Ideally I would like smart batch capabilities, emphasis on image correction and tweakable OCR. An example use case: I have photos of textbook pages - the most ideal application ever would adjust for shadows and noise, apply filters to clarify text, differentiate between text and pictures and process separately, deskew and warp to compensate for curved pages, crop correctly, provide tools for OCR adjustment, and compile into a PDF. Correct me if I'm wrong, but all of these tasks are not too difficult from a developer's standpoint. If this software doesn't exist yet, whoever does it best will make millions!
Solved, with a great app called Prizmo. A few baffling interface problems (slides occasionally rearrange themselves or change their thumbnails to adjacent pages?!) but I managed to turn a set of low-light camera phone pictures of a textbook into a flat, curvature-corrected readable pdf. OCR didn't work at all but the images were so low quality that I was surprised at the few times it did; if I were more careful with the photography it could easily have read a page taken with an iPhone camera. Really great curvature correction, only a few interface adjustments would make this software perfect and efficient for a higher number of pages, like key shortcuts and some thought put into the curvature adjustment interface, or the ability to copy curvature correction settings over similar pages, or the somewhat obvious ability to pull 2 pages out of a picture/scan of a book. If they fire their current UX lead and replace him/her with a technology user the next update will be absolute gold.