Perceptual Document Analysis Area

Intelligent Systems Laboratory

Palo Alto Research Center

---

People Papers

The PARC Perceptual Document Analysis Area is focused on identifying visual layout and appearance structure in document images in support of targeted applications including Automated classification of documents in in production and office systems Identifying fields for structured data extraction Identifying differences between PDF documents, no matter how created Finding similar and duplicate images, slides, and documents in databases Looking up product labels from cellphone images Sketch and diagram recognition on pen computers Cross-linking and organizing large document collections

Meaning in documents is conveyed not only through text content, but also through visual structure reflected in layout, fonts, graphics, diagrams, logos, and annotations.

Our goal is to develop a technology base for perceptually-enabled applications making use of image content, especially in document images.

Current and recent projects include:

• Doctype classification
• Parsing tabular layouts
• Labeling of image content by marking type (machine print, handwritten, graphics, etc.)
• Document image fingerprinting for database indexing
• Visualizing modifications between document versions

PDA's technology base is scientifically rooted in the disciplines of Perceptual Organization and Machine Learning. Document images are an especially ripe domain in which to implement the Gestalt laws and other approaches to visual perception.

One of our application projects is the ScanScribe document image editor.

The PixLabeler program for groundtruthing document images.

The sheepdog game.

---

Group Members Eugene Bart Doron Kletter Jing Lin Prateek Sarkar Eric Saund, Area Manager

Current and Past Affiliates Alex Brito Jeff Briedenbach JD Chen Jim Mahoney Asghar Nafarieh David Fleet Dan Larner Ed Lank Tom Moran Sandrine Ribeau-Sergeant Stu Claassen Todd Cass Yizhou Wang

---