Geospatial Content-Based Image Retrieval Viewing Interface

1
Geospatial Content-Based Image Retrieval Viewing
Interface

• MENTORS Dr. Chi-Ren Shyu Matt Klaric BY
• Abdulmajed Al-Meshaal Neil Timmerman

2
Topics of Discussion

• Introduction
• Old System
• Goals
• Advantages
• The new system
• Alternative Solutions
• DEMONSTRATION
• Conclusions

3
Introduction

• geospatial data research has been focusing on
  geospatial data and advanced software to improve
  computer imaging and mapping of target areas
• to try to improve the existing user interface of
  the geospatial content based image retrieval
  system developed by Dr. Shyu and Dr.Palaniappan
• The system users Department of Defense,
  construction firms, city planners, and
  researchers

4
Introduction (Cont.)

• Geospatial Content Based Image Retrieval (CBIR)
  system
• Be used to query databases of images without any
  corresponding textual information describing the
  photographs at all.
• Many CBIR systems utilize multiple channels of
  the same images to improve the results.

5
Old System
6
Goals

• Implement a user interface to the geospatial CBIR
  system that complements the CBIR software and
  database using IIP-Image
• Improve users interface allowing the users to
  use the mouse directly on the photo and scroll
  around allowing them to view several tiles
  partially in any ratio of one tile and
  corresponding tiles that they desire
• operates without the need for page refreshes
• allows users to pull up a photograph based on
  its longitude and latitude

7
Advantages

• improved, smoother, and more professional user
  interface to the geospatial CBIR.
• users to operate the same photo and to focus on
  sections they like
• save time and improve computer performance.
• viewer will be able to view small parts either
  individually or in clusters as they choose and
  can at any time zoom in or out of the photo
  without losing the image quality

8
System Diagram
9
IIPImage Client Side Interface
10
Alternative Solutions

• To leave the CGI server as is and focus on
  implementing all of our additional
  functionalities in the Java Applet itself.
• Modify he CGI server instead of modifying the
  JAVA applet. The CGI Sever contains several
  thousand lines of code and will require much more
  research in order to effectively understand the
  correct way to modify the code for our purposes

11
DEMONSTRATION
12
Future Work

• Creating and Implementing an algorithm that
  determines a zoom level, or how much area to show
  in the screen based upon the users
  latitude/longitude accuracy. In other words, if
  the user types in 157.8559, then this method
  would return an appropriate x/y area, or perhaps
  within the Applet, a zoom level that
  automatically returns a standard area.
• A way to scroll over to existing adjacent images.
  As it is now, you can scroll all over the
  currently loaded tiff, and it even caches
  previously viewed tiles, but it only plays within
  the boundaries of the one tiff. We would like it
  to check the CGI server for adjacent images, and
  load them as well in addition, for complete
  continuous scrolling.

13
Future Work (Cont.)

• Since the whole point of this interface is for a
  Content Based Image Retrieval System, obviously
  wed like to implement a way for the user to be
  able to click on a tile, and query the CBR
  database for matches. We would want to show those
  matches in a simple HTML page and allow the user
  to click on query result images and load them
  into the IIPImage Applet interface, or perhaps
  pop a new browser view with the IIPImage Applet
  and the selected image.

14
Conclusions

• More time is need when your working with open
  source software that has no documentation at all!
• The IIPImage software was not designed to switch
  images and adding that ability was like trying to
  hold it all together with chewing gum and hair
  ties. In other words, a ground up redesign, with
  that functionality in mind from the beginning
  would be much more optimal.
• The IIPImage softwares scrolling and Applet
  interface serves as a good starting point for a
  prototype, however, modifying its undocumented
  thousands of lines of code without breaking it is
  nearly impossible.
• After much frustration, it is clear IIPImage will
  not be the solution to this set of requirements.

15
MANY THANKS TO

We would like to thank Matt Klaric For
helping us in this project
16
QUESTIONS?