Mobile Retriever: access to digital documents from their physical source

In this paper, we describe an image based document retrieval system which runs on camera enabled mobile devices. "Mobile Retriever" aims to seamlessly link physical and digital documents by allowing users to snap a picture of the text of a document and retrieve its electronic version from a database. Experiments show that for a database of 100,093 pages, the correct document can be retrieved in less than 4 s at a success rate over 95%. Our system extracts token pairs from the text, to efficiently index and retrieve candidate pages using only a small portion of the image. We use token triplets that define the orientation of three corresponding tokens to effectively prune the false positives and identify the correct page to retrieve. We stress the importance of geometrical relationship between feature points and show its effectiveness in our camera based image retrieval system.