How to fix a sideways or upside-down scan

Open the scan in Read by Zaweba and it fixes the orientation for you. Read detects pages that are lying on their side or standing on their head, turns them upright and then reads the text, so you don't have to rotate anything by hand first. It works on scans, photos and scanned PDFs alike.

Sideways pages are one of the commonest problems with scanning. Someone feeds a landscape form the wrong way, a document feeder flips a sheet, or a phone photo of a page is taken while you lean over a desk. To an OCR engine, a page on its side is a jumble of shapes, and an upside down one can produce confident nonsense.

How Read decides which way is up

There are two separate questions, and Read answers them in turn. First, is the page sideways? Read looks at projection profiles, which is a way of adding up the dark pixels along rows and columns. Lines of text make a strong striped pattern in one direction and not the other, so the page's direction shows up clearly.

Second, once the lines run across, is the page the right way up or upside down? Both look like neat rows of text in a profile. Here Read does a trial read and keeps the orientation that gives sensible words. It's a practical test rather than a guess.

The same clean-up also straightens pages that are only slightly crooked, within plus or minus 15 degrees, so a page that is both sideways and tilted is corrected in one pass.

Doing it, start to finish

Open Read by Zaweba and drop the scan on the window. You can also paste it, open a folder of scans, or scan straight from a WIA scanner inside Read. For a scanned PDF, Read turns each page into a picture at 300 DPI before reading, and every page gets the same treatment.

Look at the Clean-up tab and the Document tab. Each clean-up step acts only when it's needed, and the Document tab tells you what Read did, so you'll see when a page was turned. The cleaned-up picture appears beside the text in the review, the right way up, with the words alongside it.

Then save the result the way you need it: copy the text, save a .txt or a .docx, or make a searchable PDF. In a searchable PDF the page shown is the cleaned-up picture, so the saved file is upright too, which is handy when the original scan was the wrong way round.

Phone photos are a slightly different case

Phones store an orientation tag with each photo. Read uses that tag first, so a picture taken with the phone held sideways opens the right way up straight away. The projection and trial read checks then catch anything the tag got wrong, such as a photo taken looking straight down at a desk, where the phone often can't tell which way you meant.

Getting a good result

Orientation detection needs lines of text to look at. A page that is mostly a photograph, a single short label or a big decorative heading gives it less to go on. If a page still comes out wrong, check that the scan is sharp and reasonably large, since low resolution pictures are one of the honest limits listed on Read's Help tab.

For a stack of mixed pages, use the Batch tab. Read works through a folder in the background, fixes each page's orientation as it goes, and reports any file it couldn't read without stopping the rest. You get one combined document or a set of files per picture, all upright, made by Read by Zaweba on your own PC.