CS6825: Computer Vision word cloud

Digital Libraries: Content-Based Retrieval

One important area of interest in Computer Imaging is the storage and efficient retrieval of images in a database or "digital library". How can we retrieve images? How can we describe them? How can we store so many images?

Content-Based Retrieval

  • Describe data by its contents.
  • Can be used for digital text, sound, music, image, video, e
  • Works like a search or a browsing tool -- analogous to the current web search
  • Traditional keyword indexing is fast and easy but, has limitations:
    • How do you say find me an image of a forrest?
    • Keywords are often inaccurate.

Current possibilites to search based on:

  • Histogram based measures (color or intensity).
  • Edge-feature information.
  • Motion Queues
  • Texture measures.
  • Features merged with keywords.
  • Knowledge based (geometric models, fiducials).

Other issues:

  • Capitalize upone multiresolution approaches
  • Work in the compressed domain to save time.



Real Systems

  1. IBM's QBIC System (Querry By Image Content)
  2. Virage System: Visual Information Retrieval Imaging System

 

Some Papers through 99

© Lynne Grewe