OCR is required, and there are two paths to follow:
1. OCR DLL (including image pre-processing because that's required for a good success ratio anyway, IMHO) developed by the coder.
2. Image processing DLL + 3rd-party library (suggested by coder), as long as the combined libraries meed the general conditions (Win32, unmanaged code, minimum dependencies, thread safety, small memory requirements, no memory leaks and of course, your required success ratio).