Content conversion and content verification possibilites

Hi Ravisan,

To expand upon what Abambo has said, there are things you can do during the original scan that can help in the OCR process. You are absolutely correct to be saving them as TIF documents as opposed to JPGs. Image degradation caused by JPG creation can significantly affect the final product of document scans.

In addition, higher resolution scans are also worth the effort. A minimum 300 ppi is always good and if your scanner can do it, 600 is even better. Also, if your document has smaller fonts, than 600 ppi is almost mandatory. The reason for this is that certain letter combinations can confuse the OCR process if (1) the scan is a low resolution scan, (2) if the scan is not a clean scan and (3) if the font in the document is very small. An example of letter combinations that can cause issues include the letter "ri." This can be seen by OCR is an "n."

Another big issue is if you were to place your pages on a scanner, hit the scan button, and say you are done. You are probably not done if you want a clean scan. What I mean by this is more easily explained by looking at a blog I wrote for Adobe here:

https://forums.adobe.com/community/creativepipeline/blog/2018/01/22/scanning-clean-search-able-pdfs

One last comment/warning are hyphenated words at the end of a text line. OCR is only recognizing shapes, not words. Thus if you have a text line where the last partial word is "par-" and the next line continues with "tial," you now have two misspelled words where in fact, there is no misspelled word.

If you do not currently have Acrobat Pro DC, I suggest you download a copy and try it during the trial period and see if it satisfies your needs.

As far as your desire to have a pixel by pixel comparison of the before and after documents, about the only thing I can think of at that level is to open the before and after into Photoshop, and place them as layers into the same document. Then set the top layer's blending mode to "Difference" and any altered pixels will then stand out. Then you have to zoom into each document looking for potential light-shaded pixels.

There are two ways you can look for errors. One is to save each document out as Word files and let Word underline any spelling errors. While this lets you see the entire document very easily, it also prevents you from making any corrections IN the document.

Alternatively, in Acrobat Pro DC, select the Enhance Scans tab and on the top, below the Recognize Text, dropdown menu, select Correct Recognized Text.

This will highlight all of the words that are questionable. Please note that the only two actual mistakes were in the shown section included the word "instruction" and lower down, after the text All-Natural Hand Cleaner there is a small graphic of "USA" also was seen as an error. All of the other selected text items were actually spelled correctly.

Important Note: this will not show how the word was misinterpreted as the OCRed text is invisible. Fortunately at the top of this screen you do see a closeup of the scanned text and how the text was interpreted as seen below.

This was an advertising flyer I happened to have on my desk I grabbed to demo this. The font was small and the scan was done at 300 ppi.

Let me also add that I redid the same scan but teased the Levels (or Histogram, same thing, different name) and improved it (I enhanced the contrast between the page and the text) and this improved the accuracy considerably: there were no actual errors at the same 300 ppi (although Acrobat did want verification on some of the items).

Please let us know if this helps your decisions.

Discussion successfully moved from Forum comments to Creating PDFs

Converting Tiffs to PDF: Drop them into Acrobat...

Checking: I do not see the use of such a tool and I do not think that there is one. If you don't do destructive compression or data modification (like colour model change, size change) tiff data in a PDF will not differ from tiff data in the tiff file.