@mouser: Many thanks for this info!
I'm glad
@dvally mentioned this bug as I had been unaware that SC had "fine rotation" of images. I shall use that now I know it's there! I would otherwise normally have used the fine rotation of images feature in
Picasa.
I regard images containing text as being
data, and for straightening up
images containing text and extracting the data, I would normally use a different approach:
- ScanTailor: I use this on book/document scans (after reading about it elsewhere in DCF). ST is great for straightening/levelling up text and then OCR scanning it. For example, ST automatically splits up pages of two-page images of a book, straightens them, cleans up any dirt or image noise and then feeds it into an OCR scan. It does a superb job - perfect, in fact. Very good for minimising OCR scanning errors.
- OfficeLens: (A free smartphone app from Microsoft, for Android and Windows OS.) I have been trialling it in a Nokia Lumia 830 for capturing text from images of documents on a desk, images of whiteboards, images of business cards, and any other images containing text (e.g., advertising text on the sides of a panel van). You feed the image (as captured by the smartphone camera), into the app., which then automatically tidies it up and removes any slanting perspective. The image doesn't have to be taken square-on or anything, as the app figures it out, squares it up and also removes glare on the object (e.g., whiteboard reflection). The app can then save the tidied-up image to MS Office OneNote, which collects and indexes any text (captured by OCR) in the image. In the case of business cards, the app cuts out anything which is beyond the edges of the card, squares it up and passes it to OneNote with an associated panel of OCRed data that is extracted from the card and which is logically slotted into fields - e.g., "email" if it is an email address, or "phone" if it is a phone number, "address" if it is an address, etc. It's pretty smart (or maybe OneNote does those smart bits as post-processing, I don't know).
The reason I mention these things here is that it is an opportunity to urge consideration as to how SC (and CHS) might be able to be configured to (say) automatically and intelligently identify any
text in an image and maybe then pass it to something like ScanTailor or OneNote for OCR post-processing. I think that could be a simpler and more doable approach than re-inventing the wheel and trying to build that same post-processing functionality
into SC or CHS.
