2011年6月28日星期二

3 iPhone optical character recognition apps compared

If you're going paperless by transferring all your printed documents to digital text or PDF files, using OCR (Optical Character Recognition) makes the job much easier by eliminating manual input. Luckily, you can use your iPhone to do OCR, eliminating the need for any additional equipment.

The three apps I've tested here are Prizmo, Perfect OCR: document scanner with high quality OCR and Image To Text – OCR. Only Image To Text is free; Prizmo is $9.99; and Perfect OCR is $3.99. I based my evaluation of each on three criteria: results, interface and speed. The text I used to test the apps was Apple's "Here's to the crazy ones" quote.

Results

With plain, printed text, set in Helvetica on basic white paper, the best results came from Image To Text. It produced no errors in the generated digital text, while Prizmo and Perfect OCR (especially Perfect OCR) did produce mistakes. This is probably because Image To Text performs the OCR process on its server, rather than on-device like the other two apps, which means it can take advantage of much more powerful processing resources. You can see the results in the gallery at the end of this post.

On the second test, using a more cursive font, Corsiva Hebrew, Perfect OCR didn't produce any results, since it couldn't recognise any text in the image. Again, Image to Text got the best results. None of the apps generated results anywhere close to the original text, but Prizmo didn't produce any recognizable words at all, whereas ITT did.

The final test I carried out was on handwritten text. OCR isn't meant for handwritten text, but I thought I'd see what happened. None of the three apps recognized text in the image, so you'll have to look at handwriting recognition software if that's your thing.

Category winner: Image To Text.

Speed

If you have hundreds or even thousands of documents to reproduce, the speed at which your app can generate text is important. Perfect OCR was the slowest overall, since it requires you to capture three images of the text. It also took a while to process the images. Image To Text, as I mentioned previously, performs the hard work remotely, so it works a bit slower while it communicates with the server. Prizmo was the fastest of all the apps, but its speed still doesn't match computer-based software.

Category winner: Prizmo.

Interface

While not as important as speed or results, if you use an app often, its interface makes a difference.

The worst app was Image To Text. The interface is simple, but maybe too simple. You can either take a picture or use one from your library, but once you've chosen or taken the picture, you can only email results or send them to Evernote. The app is basically just a slim front-end for the real action, which takes place on the remote server. You can't crop the image you take, or edit it in any way, so often the results include text you didn't want.

Pefect OCR has a better interface, with a few editing tools, and the app stores a history of the text generated. You can crop, rotate and recolour a taken image. This means you can try to improve your results. There are also more options for sharing the text, such as copying it, and emailing it in different formats. The interface isn't always pretty, but it offers some flexible functionality.

By far the best interface belongs to Prizmo. It looks professional, has a lot of features, but isn't complicated to use. When taking a photo, you can choose from a list of possible sources, and your choice dictates how Prizmo optimizes the results. There's a grid overlaid on the camera view, and there's even a speech recognition feature which allows you to tell your phone to take a photo to prevent shutter shake. Once you have a photo, you can rotate and crop it, move it around to compensate for perspective, and even pick a point on the image to use as the measure of what constitutes the background color for the text. After the results have been generated, you can share them to Dropbox, Evernote, CloudApp and more. You can also use Google Translate to translate it to another language.

Category Winner: Prizmo.

Overall

If you don't mind sending your text off to a server to get results, and only being able to use email or Evernote for sharing, then I'd recommend Image To Text. However, if you need extra features like editing, a history feature and better sharing, then Prizmo is your best bet, even with the $10 price tag.

Loading
NextPrevious
Picture 1 of 19
Perfect OCR Cropping

Related content from GigaOM Pro (subscription req'd):






沒有留言: