Mobile Barcodes? Why Not Snap Images Instead…

In my quest to find new and innovative ways to utilize a mobile device for just about anything, I’ve always been intrigued with the idea of using image recognition technologies with mobile devices to interact with the things around you in everyday situations. Unlike mobile...

809 0
809 0

In my quest to find new and innovative ways to utilize a mobile device for just about anything, I’ve always been intrigued with the idea of using image recognition technologies with mobile devices to interact with the things around you in everyday situations. Unlike mobile barcodes, image recognition doesn’t require any special applications, or device features to operate, it can work with any mobile device made that has a camera, If done correctly.

SnapTell claims to have solved many problems faced when trying to recognize images taken with camera phones. Things like blurred images, and shaky hands can produce images that are hard to recognize. Since SnapTell works by taking images sent to the platform and comparing it against it’s central database to return relevant content, it’s essential that the image is readable even though it might not be of the best quality. How does it work? Their website says they use….

“…a highly accurate and robust algorithm for image matching that we call ASG. Image matching is the problem of efficiently matching a query camera phone image against a database of images. Our technology offers unprecedented scale in detecting a matching image in a large database of images. Scaling of image matching is achieved using patent pending indexing techniques to organize all the features in any of a database of images for the purpose of efficient lookup…”

They claim their technology will work on almost all camera-phones made world-wide, including those on the lower end of the market that feature a VGA camera and offer low-resolution photos of 640×480 and lower. It also works in real-world conditions where there might be a lot of issues including lighting artifacts, focus blur, motion blur, perspective distortion and incomplete overlap with the database image. They claim photos of magazine print ads, outdoor billboards, posters, product packaging, branded cans, and bottles and logos can all be used with their solution with no problem.

The coolest thing they offer, and the reason I found it noteworthy was the fact that they offer a “Text Extraction” solution as well that will actually pick text out of images taken with camera phones and use that text to initiate a search query. If someone takes and sends a photo of an image that isn’t in their database, it can automatically extract any text and use it for a search. It can then still return relevant content about that image without it even being in the database. Pretty cool.

In this article