Google released an Optical Character Recognition API (i.e. image-to-text) as part of the Google Docs API. I can't get it to work on the sample images I tried (I always get "Error processing document: Unable to read response, or response is empty"), though my sample docs weren't very high quality to begin with either.

► Sample app: http://googlecodesamples.com/docs/php/ocr.php
► Official blog post: http://googledataapis.blogspot.com/2009/09/import-scans-or-go-multilingual.html

Also see http://weocr.ocrgrid.org :
<<WeOCR is a platform for Web-enabled OCR (Optical Character Reader/Recognition) systems that enables people to use character recognition over networks. A WeOCR server receives document images from users, recognize texts in the images, and return recognition results to the users. WeOCR does not have its own character recognition engine. Instead, it is intended to accommodate various character recognition engines. WeOCR provides a simplified user interface so that more people can benefit from OCR easily.>>

[Via http://waxy.org/links -> http://googlesystem.blogspot.com/2009/09/google-docs-ocr.html]