Semantic Decomposition of Bitmap Images Using CHREST

Abstract

This report results from a contract tasking University of Hertfordshire as follows: The Grantee will investigate the use and modification of Chunk Hierarchy and Retrieval Structures (CHREST) a computational model of the human perceptual and attention system as a useful and powerful tool for image analysis from complex photographic-quality images. The CHREST model includes detailed processes for learning new information and retrieving familiar patterns. CHREST has been shown to capture the detailed perceptual knowledge acquired by experts in high-level domains such as chess or computer programming. This project will improve on the current system by extending it to identify and integrate components of complex images. There are three main objectives to this research programme: 1. Improve the acquisition and use of semantic categories to include arbitrary geometric relationships and more abstract relations such as inside outside' above' or below'. 2. Develop an efficient clustering technique for bitmaps to take advantage of natural generalizations in bitmap representations, so as to locate the most effective for the target application. 3. Develop a flexible user-interface for image analysis with CHREST for carrying out semantically-based image analysis with CHREST, and use it to test the model and compare its performance with human data.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 18, 2005
Accession Number
ADA430629

Entities

People

  • Peter C. Lane

Organizations

  • University of Hertfordshire

Tags

Communities of Interest

  • Biomedical

DTIC Thesaurus Topics

  • Abstracts
  • Acquisition
  • Character Recognition
  • Classification
  • Cognitive Systems Engineering
  • Data Displays
  • Decomposition
  • Eye Movements
  • Feature Extraction
  • Hierarchies
  • Human-Computer Interaction
  • Learning
  • Perception
  • Recognition
  • Universities
  • User Interface
  • Word Recognition

Fields of Study

  • Computer science

Readers

  • Database Systems and Applications
  • Neural Network Machine Learning.
  • Space/Atmospheric Physics.