Thread: OCR: what software and what settings?

I have a lot of docs to be scannedand put in PDF format.

The aim is either a PDF file(which is nothing more than a package of several JPG) or a PDF imagesearch (an invisible text is hidden and can use a search engine) orfull text PDF file.

Strangely, I found nowhere OCR softwaretesting. Other topics of HFR are a little older. I mostly scandocuments for archiving and transmission of legal documents via ftp.

I tried to do OCR with Adobe Acrobat 7 and my scanner (an oldAgfa 1212 and a recent Epson 2480) and I'm not really satisfied withthe results. I'm looking for the best compromise "respect of theoriginal document / file size."

I experimented withdifferent resolutions and it is always disgusting (even by increasingthe resolution a lot, I sometimes have strange effects), that is tosay that I often find myself with a moitiée sentence remains inbitmap, the text aute moitiée but with different fonts, etc.

I am also in the same config. I'm new to OCR, but I have a lot of legal doc scanner to archive.

On soft drinks, I agree: Acrobat 7 is deplorable, I do not like Omnipage, ReadIris remains my favorite (I have version 8 but it is not very different).

My solution depends on the quality of the original doc: there is not the same treatment if it is an old fax or original brand new well-ventilated area. I find the best value weight with "gray 8 bit" and 150dpi: OK to print, good for the screen, acceptable to the OCR. However, sometimes you go through an image processing software (like photoshop) to enhance contrast and increase the brightness a bit (so as not to have transparency in the contents of the back of the page).

Regarding research, two solutions exist: you can create a pdf "image" and then use the properties of the document and a summary of the text in the "subject" or "keywords" that allows you to make copies of the original (ie if a legal value allowing you to destroy the original paper) and keep the possibilities of access to the main content.

The choice is bottomless GIF at the end, when the final export. Currently I tend to scan via TWAIN since ReadIris gray 300dpi scanner and all the pages one after the other and then save the TIF file (containing all the pages), brightness contrast before any editing imported back into Readiris.

There I proceeded to the recognition of the document and then export to Word XP, lightweight design (it affects Title 1, Title 2 and Title 3 standard doc) then print directly to Acrobat 7. Here I get the text file with full bookmarks (Title 1, Title 2, Title 3 ...) in the tree. I resume the file containing the TIFF images in Photoshop, 150dpi gif degradation, and backup. In Acrobat, inserting page after the current page, gif. It remains only to delete temporary files, finished.

In fact you can not save a single file in multiple pages and have them recognized by photoshop. We must therefore save each page of the document in an individual file.

For nothing prevents against then use the batch function in Photoshop to apply the same treatment to all files.

Hello everyone,

I need to create a batch file that runs at night on a server for a java application to convert a PDF "image" in a PDF "text". Given that I do not have "full" version of the software above, does anyone just tell me if this is possible with one of them. I know quite well but I FineReader tested a "light" version, not pro.

The principle is:
I must load into blobs oracle pdf files from scanned documents by users in order to make search text. But each user does not necessarily use the same software with their scanners , So necessarily does not get a pdf with the ability to search.

I am currently looking for a powerful OCR software and unlimited number of pages (when Abby FineReader).

Omnipage, Autobahn is not bad but far, far from perfect in terms of recognition (despite a TIFF 600 dpi, 256 gray scale input).

Do you have any feedback on the software work?