HP Labs India

In emerging markets many enterprises and businesses have a number of paper based processes. Our work in this project aims to do automatic document repurposing, obtain the value of IT even for paper based processes and look at problems related to document archival and retrieval.

Our work focuses on a range of sub-topics such as: page segmentation and classification, document image registration and ink extraction, document image matching and comparison, document retrieval from images and also basic document image cleanup targeted at HP’s imaging devices.

Handwritten annotations are often used in offices and business processes. While moving from the digital world to the paper world is relatively easy moving the other way around for capturing handwritten information and annotations is a hard problem.

Our work focuses on intelligently segregating the ink from the paper document and recognizing it accurately. How can you distinguish handwritten text from printed text effectively? Also how well can you recognize the extracted handwritten ink? While OCR of printed text for the Roman script has achieved high-level of performance, there is still substantial work required for handwritten text.

Enterprises and businesses have significant problems related to document forgery and fraud. For the issuing enterprise it is a question of credibility and also the costs involved in verifying document. For the verifying enterprise it is the problem of loss due to fraudulent documents, costs involved in verifying these documents and also the overall slowing down of business transactions.

In this project we deal with innovation and technologies that can help in the prevention of document fraud within the existing ecosystem, support for automatic authentication of paper documents, etc. Our work on Document Authentication System is related to this. We have also looked at techniques that can distinguish copies from originals.

Our research focus is around security techniques that can prevent and detect document fraud and increasing the density of machine related content on paper documents.