The problems of recognizing blobs (the result of connected component analysis), classifying them as representing potential collections of interest such as text strings, and applying techniques such as are described in the patent application, have been known and practised for decades. In fact, if faced with this problem, it's impossible not to do what these gentlemen describe, and wish to patent, or a close facsimile thereof.
As mentioned in a comment, this is a patent application that as of today (Feb. 20, 2013) has not yet been examined. All of what it describes is not what it desires to claim and the broadest claims found in a published patent application are usually narrowed in the course of prosecution. As published the claim 1 is:
Since this is published but not yet subject to an office action on the merits it is in the time window allowing 3rd parties to submit prior art into the case. The mechanics for this are on several other Ask Patents posts. This is the USPTO press release on the topic. Up to three documents per submitter, with limited comments, can be filed at no cost. It is possible to do yourself. If done through a registered practitioner it can even be done with anonymity.
Please look up this application in USPTO Public PAIR to see that the applicant themselves have already made the examiner aware of several prior art patent documents.