EPSRC logo

Details of Grant 

EPSRC Reference: GR/H48248/01
Title: AN INTELLIGENT SYSTEM FOR CONVERTING SCANNED DOCUMENTS TO SYMBOLIC FORM
Principal Investigator: Xydeas, Professor C
Other Investigators:
Oakley, Dr JP Oakley, Dr J
Researcher Co-Investigators:
Project Partners:
Department: Electrical Engineering
Organisation: Victoria University of Manchester, The
Scheme: Standard Research (Pre-FEC)
Starts: 08 July 1992 Ends: 07 July 1995 Value (£): 99,884
EPSRC Research Topic Classifications:
Image & Vision Computing
EPSRC Industrial Sector Classifications:
Creative Industries
Related Grants:
Panel History:  
Summary on Grant Application Form
The primary objective of this work is to develop a flexible system for converting complex, scanned document pages into an editable, electronic representation. The required page complexity is such that typical office documents, such as those defined by the ODA model application profile Q112, can be processed. Furthermore, we intend to investigate ways of generalising the scope of operation of the system using interactive training processes.Progress:A powerful and flexible Intelligent Document Understanding (IDU) Architecture has been developed and a prototype system has been implemented which can determine Page Layout Structure and identify different types of Layout Objects within the page. The advantages of our system are: Robustness to noise and distortions via use of multiple image analysis algorithms. Flexible hybrid control architecture implying good generalisation characteristics. Use of content information derived from text recognition as an evidence source.The work modules completed to date comprise: Design of custom image analysis library. Design of text recognition routines. Representation of document layout structure using a frame based model. Design of a generic Forward Production System for bottom-up Layout Object detection.The prototype IDU system has been tested on a specific technical journal (the Image, Vision and Computing Journal). This document contains pages which are of a complexity associated with the ODA application profile, Q112 and contain a mixture of text, equations, line-art and photographics in a two column format. Research has focused on ways of matching Layout Objects, detected by the FPS, to the document model. This problem has been shown to lead to combinatorial explosion, making exhaustive search impractical. To circumvent this, a multistage search procedure has been developed which is computationally inexpensive without losing matching accuracy. We have also investigated two matching criteria: one based on maximal area coverage and one based on co-occurrence probabilities of the defined Layout Objects. To quantify the performance of the IDU system, we have defined two metrics: the Object Identification Rate (OIR) and the Percentage of Column area (PCA) correctly interpreted. System tests using a 50 image test set indicate a performance of OIR of 96.1% and a PCA of 95.2%.Future investigative work will focus on ways of generalising the operation of the IDU system. In this respect, we will investigate possibilities for both automated and interactive learning.
Key Findings
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Potential use in non-academic contexts
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Impacts
Description This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Summary
Date Materialised
Sectors submitted by the Researcher
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Project URL:  
Further Information:  
Organisation Website: