Paper-Based Augmented Reality



  Paper-Based Augmented Reality.docx (Size: 866.52 KB / Downloads: 0)


INTRODUCTION

Linking the physical and digital worlds is a long standing goal of augmented reality
Previous augmented reality techniques for paper documents alter their appearance with bar codes or textured paper. Often, a special purpose device, such as a pen with a camera in it, must be employed to recognize the embedded code.
These characteristics all inhibit the use of augmented reality with paper documents.
Our prior work indicated that images of small patches of text contain enough information to make them as unique a s a finger print . We showed that it was possible to distinguish a small rectangular region (one inch square) from among thousands of other text image
patches. At that time, we leveraged this characteristic to identify the electronic original for a given paper document. However, the same result indicates that patches of text can be used as markers in an augmented reality system in which arbitrary x-y positions in printed
text passages can be associated with electronic data.



ALGORITHM OUTLINE

The operation of a paper-based augmented reality system is illustrated in Fig. 1. PBAR-enabled documents are created by scanning a document and indexing it for text patch recognition as shown in Fig. 1 (a). Data is associated with regions on the document by choosing “hot spots” that are rectangular patches of text, and adding data to the hot spots (Fig. 1 (b)).
The index information and symbolic hot spot data are stored in the PBAR database. A simple example of data in a hotspot is a URL that points to a web page. However, it could just as easily be a video file, audio clip or even an electronic version of the original document itself.


TEXT PATCH RECOGNITION

The objective of the text patch recognition algorithm is to correctly determine the identity of a page and the x-y position in the page of a small patch of text. The technical challenge is illustrated by the image in Fig. 2 that shows the typical quality of images produced by
Commonly available camera phones. Characters are so blurry that “OCR” is basically impossible. However, it is still possible in almost every case to identify the bounding boxes around words since the spaces between words and lines can still be distinguished.



APPLICATIONS

There are many possible applications for PBAR. Important considerations include whether the PBAR database is on the phone or on a server and whether the database is created as a side effect of printing a document on a PC.
This section presents two examples,
• Travel Guidebook
• Self Printed Documents



SELF-PRINTED DOCUMENTS

Documents that are printed on a desktop PC are typically created by individuals for their personal use. PBAR allows users to customize the interactivity of those documents based on their own needs and permits the database to be under the user’s personal control: it could be shared on a networked server, saved on the individual’s PC, or pushed to the user’s camera phone.
We created the architecture for printing and indexing web pages shown in Fig. 7 that automatically captures an image of every document in the print driver, indexes them in the PBAR database, and associates the URL’s in the document with their physical location on the web page. This includes a plug-in in Internet Explorer that exports URL’s to text patch feature extraction software. This system is fully implemented on Windows XP and
Vista and can be installed on any PC.