Optical Character Recognition (also known as OCR) is a technology that allows computers to convert bitmaps (pictures) to ASCII text.

With the emphasis on ''character'' recognition, OCR of printed text is rather accurate (as of 2010), but not very useful yet.  This is because it is not able to understand things beyond the character.  It is quite common for modern documents to utilize a variety of fonts, angles, borders, backgrounds, columns, tables, and inline images which current OCR engines will choke on.

To deal with this, you can first pass the image through a layout analysis engine, which attempts to break the image up into areas and classify them as text areas or 'other' areas.  Only the text areas are then passed through OCR.  After OCRing the text areas, the document can be reconstructed in a different format by piecing together the recognized text and the 'other' areas.  

Layout analysis may be much more complex than OCR.

----

The AntiCaptcha