http://code.google.com/p/tesseract-ocr/ has some wrapper to use it in .NET, or, simpler:
http://www.codeproject.com/KB/office/modi.aspx but you need to keep an eye to the license since it is a part of the Office suite. In both case you tipically need some pre processing for the image and, as a solution I did in the past, some post processors that using some ehuristict correct the mistaked words.
