A way to undo Formatted Text & Graphics OCR from Acrobat 7?

Question

Over the course of a few months, my company received a large number of PDF files for a project for which the internal policy was that every file should be text searchable.  Unfortunately, we did not save the native files in any sort of convenient way, having at that time not realized that failing to do so was a very bad idea.  We ran OCR on every one of the files that we received, which total approximately 4,000.  At the time that we received the majority of these files, my company was still using Acrobat 7; we've since upgraded to version 8.
Recently we discovered that there were discrepancies between our electronic copies and the hard copy printouts from which our electronic copies had been generated:  in the electronic copies, uppercase F had changed to P, S had changed to 8, etc.  We eventually worked out that it must have been that at some point a computer was mistakenly set to run OCR using the Formatted Text & Graphics setting, as opposed to either Searchable Image or Searchable Image (Exact).  This was absolutely not want we wanted, as for our purposes using a type of OCR that causes the original images to change essentially renders the files useless.  My questions, then, are the following:
1)  As I asked in the title, is there any way of undoing Formatted Text & Graphics OCR that was performed in Acrobat 7?
2)  Is there a way of identifying files that have had Formatted Tex & Graphics OCR performed on them (something stored in the metadata)?
Rebuilding these files from scratch is going to require a gargantuan effort, so any help would be much appreciated.

Answer 1

Hi,
Bernd's been across the mountain and seen the bear; so, you can bank on what he posted.
But, just because, I'll second his "no".
Formatted Text and Graphics (Acrobat 7, 8) and ClearScan (Acrobat 9, X) effectively replace the image of textual characters.
If a character is not recognized as 'something' a bit map is of the thing is left behind.
Now, while Acrobat or other OCR engines (Abbey FineReader, AdLib, Adobe Capture, etc.) are really rather impressive no OCR engine has 100% accuracy 100% of the time. Other variables  come into play (scan lamp age/brightness, platen cleanliness, scanner mechanicals cleanliness, calibration of scanner, hard copy 'quality' (characters' darkness density, contrast between characters and background, presence of lack thereof of boxed in text, text in or adjacent to line arcs/circles, etc.).
All of that is for semantic content that is "textual". Semantic content that is not textual (but, coincidently may contain text) provides little to no useful OCR output (e.g., graphs, drawings, etc.). Validate this by performing OCR on such a PDF then Export to a plain text file. Print this file out and compare that to the source paper or the scanned image.
There is no metadata info that identifies the OCR mode used.
Perhaps something buried in the bowls of PDF page description content; if so, not intrinsically easy to obtain.
My suggestion (fwiw) - move forward with re-scan.
A server product would help to move it along but a high speed scanner hooked to a local machine (with ample resources) and Acrobat Pro 8 or 9 get it done. With Acrobat 8 or 9 use Search Image (Exact).  In Preferences check the category Create PDF or TIFF to assure it is what you desire. Check Acrobat's scan presets to assure you have what you want vis-a-vis Compression and Filtering. Do avoid "Automatic".
Be well...