What if you could scan documents using a digital camera instead of a flatbed scanner? What if you could “scan” all sorts of documents, including billboards, rare and antiquated letters, treaties on parchment, and even rare books, without ever touching the surface? If you could do that, you would be a reprographic photographer. Reprography is a specialised job, with specialised and often very expensive equipment, but Prizmo gets close enough to at least offer entry-level reprographic capabilities for the price of three pints of bitter.Prizmo 1.1, a Creaceed application, has some nifty features. Judging by its simple interface, it looks just a lightweight image editor, but under the hood Prizmo 1.1 has an image processing engine that enables very skewed images to be straightened without sacrificing detail, and an OCR engine. The combination of the two in theory allows you to take a photograph of whatever document—invoice, letter, cashier ticket, billboard, book…—straighten it and OCR its contents.
This in turn allows such documents to be indexed by Spotlight. It also allows you to digitise your office and your entire administrative life by “scanning” all of your documents using nothing but the digital camera you’re sure to own (from mobile phone to dSLR). This also was the initial reason for developing Prizmo, but as so often happens users are finding new ways to deploy Prizmo—and they are constantly moving the boundaries of what they want Prizmo to do.
So, while Creaceed started with Prizmo as a replacement for a scanner, Prizmo users are now asking for the ability reproduce rare books, which challenges Prizmo’s engine to a point that an update will be required to accommodate for the waviness of a book page when it is not forced to lay flat down.
To be honest, I was sceptical of Prizmo’s capabilities. I didn’t doubt Creaceed’s ability to get the image engine right—after all, this is the company that got us MorphAge Pro, the morphing application that even colleges use to check for genealogical resemblance, etc. I was sceptical of Prizmo’s OCR capabilities.
So, of course I tested Prizmo 1.1 with a number of documents. Using Prizmo is easy enough: just drag an image to the window, crop it, create a calibration of your lens or use the application as is, and let it correct for the document’s skewness and other faults. Prizmo will straighten the text in the cropped area and you will be able to read it on-screen. However, at that point in time, it will only be an image. No OCR applied whatsoever.
If you want OCR, you’ll press the OCR button. Creaceed integrated the technology of OCR specialist ExperExchange into Prizmo. Prizmo’s character recognition technology supports 10 languages (English, French, German, Dutch, Italian, Spanish, Portuguese, Swedish, Danish and Norwegian). And the results are Spotlight indexable.
OCR results mixed but unaffected by image skewI found the OCR results to be good on average, with some texts recognised with a 99% accuracy and others below that. OCR is something of black art, no matter whose engine you’re using. The interface to get to the OCR part, though, was less to my liking. The way it works is that you click on the OCR button after which you select your language and move the Binarization slider around. However, on my old machine, more often than not, nothing seemed to happen.
When it did work, the text became either better readable or barely readable. The art is too select the most readable setting of the slider. If that succeeds, you can hit the Recognise button and Prizmo 1.1 will OCR your text. The process is not fed back to the user. The only way you know OCr has finished is when the paragraphs have dotted lines around them. Hovering with the mouse over the paragraphs also shows you the recognised text.
The image with or without OCR result can then be saved to disk. As I said before, the results of the OCR engine were mixed, which I expected, but my main criticism here is that in my opinion, you don’t get enough feedback of the process. Even with mixed results, though, I must admit that document skew did not seem to have an influence on the OCR engine, so that’s good, because that means the skewness compensation inside Prizmo is so good the OCR engine doesn’t seem to notice any skewness at all.
Prizmo 1.1 is not yet capable of compensating for wavy text such as the text distortion you’ll get when trying to scan a book page that you haven’t flattened by force—which of course you can’t do with expensive or rare books. Nevertheless, I was capable of getting decent results with moderate “waviness” such as curled paper (from moist)!
Bottom line: Prizmo is ridiculously cheap for a product that already has powerful features and which promises so much more. I’d say give the demo a go and see for yourself!