Scan Like A Pro

Several years ago I read an article on scanning personal paperwork and reducing physical clutter. I liked the concept (except the part about using perforce!) and have since refined my own workflow to achieve something similar.

These days I scan and OCR nearly all paperwork I deal with (bills, contracts, letters, statements, etc), store them in a shallow directory structure and commit them to a git repository.

Step One: Scanner

For scanning, I use a document scanner at a clients office. I’m there a couple of days a week, so every now and again I take a pile of paper in and feed it through. They have a Xerox DocuCentre-II C2200

• it can scan 75 double sided pages at a time and saves PDFs to a windows share. Depending on the document, I scan in colour or greyscale at 600 dpi - the saved files are larger, but I deal with that later.

Step Two: OCR

The resulting PDFs have a single scanned image per page - I really wanted to OCR them if possible and make the content searchable. I explored various open source options that could run on my Linux systems, but none worked as well as I hoped. They generally had accuracy issues or would only output the recognised characters to a text file.

A few years ago I was working at a University, so I splashed out on an academic copy of Adobe Acrobat Pro 9.0 for Windows. Dual booting is a pain, but the OCR support is second to none and makes the hassle worthwhile.

Version 9.0 or later of Acrobat Pro has an OCR feature called ClearScan. After recognising the characters on each page, it converts them to a custom vector font and replaces the scanned image of that character with the vector character. This makes the text indexable by standard PDF search tools (like spotlight, etc) and has a massive impact on the file size. A 24 page text-only document I scanned last week went from 11Mb to 700Kb.

To use ClearScan open the scanned PDF Acrobat, select the Document menu, then choose “OCR Text Recognition” and “Recognise text using OCR”. Check the settings in the window that pops up and use the edit button to ensure “PDF output style” is set to Clearscan and “Downsample” is set to Lowest (600 dpi).

Step Three: Git

As a final step I store the files in a simple directory structure and save them to a git repository. A version control system might seem a bit odd, as the files generally don’t change much (if at all) but git makes it easy to store a backup of all the files on my server.

I group the documents by type and year. Anything more complex is unnecessary as I usually find what I need by searching. Here’s a sample of the directory structure:

    banking/2010/20100630 - statement.pdf
    banking/2010/20100720 - letter.pdf
    banking/2010/20100731 - statement.pdf
    insurance/2009/20090317 - contents renewal.pdf
    insurance/2010/20100315 - contents renewal.pdf
    receipts/2010/20100915 - laptop.pdf

Step Four: Profit

It has proven to be incredibly useful to have all my documents scanned and searchable. Not only is there less stuff floating around my house, but looking up records has become a trivial task.

I regularly look for files to send to my accountant and to check when I can expect a new bill to arrive. Recently I also had to look up receipts as part of an insurance claim.

Now that I have a standard workflow setup, the whole process is relatively little work and the results are convenient.