Unless you are a Ph.D. level computer scientist with years of time to spend on the problem, I'd recommend you be awestruck by the challenge inherent in Arabic OCR, and, assuming you don't have the financial resources to buy one of the very expensive commercial libraries that enable Arabic OCR for .NET (like LeadTools), you look at Tesseract, which is open-source, and which does support Arabic as of version 3.01: [
^].
By going to this page: [
^] and searching for "Arabic" you can read the history of what features related to Arabic OCR have been added to Tesseract.
I have not used Tesseract myself.