Distributed Image Reformation -- Imaging Lockheed Martin
Sponsor Information:

Bill A. Rawlings
Lockheed Martin M&DS Reconnaissance Systems
Applications Software Engineering Manager

Phone # (623)925-7574
Email: bill.a.rawlings@lmco.com

(Powerpoint slides that David Ralls used to present this project)

Project Overview:

Problem:

The end product of Synthetic Aperture Radar (SAR) processing is a byte-stream formatted image. Due to the resolution of SAR imagery, these products tend to be quite large, sometimes on the order of several gigabytes. This byte-stream imagery is not supported by commercial web-browser software for display, while out major customers desire more capability for web-based examination and distribution of SAR imagery. Our current technology is to use a monolithic algorithm to convert the byte-stream imagery to .TIF format. From there, we convert the .TIF to .JPEG or .GIF file formats for display within web-browser software. Due to the large amount of data inherent in a SAR imagery file, this image reformation process is very expensive in both time and hardware resources. We have recently encountered several cases where it has taken longer to reformat the output imagery than to produce it in the first place.
 
Solution:

A viable solution to the image reformation problem would be to implement the reformation algorithm in a distributed hardware/software environment. In this architecture, the byte-stream image would be parsed out into tiles or rows, and these tiles or rows sent to individual services running on different hardware platforms connected to the network. Each service would reformat a tile or row, then retrieve another tile or row to work on, until all tiles or rows comprising the image have been reformatted. This architecture would allow image reformation to be performed in a time effective manner on a network of low-cost Windows or Solaris platforms, and free our expensive super-computers from performing this task.
 
Implementation:

In 1999, Sun Microsystems released the first version of Jini (pronounced "genie"). Jini is an infrastructure designed to make building distributed systems more feasible. Using Jini technology with the Java programming language allows the design of truly hardware agnostic distributed processing systems. One of the services within the Jini infrastructure is JavaSpaces. JavaSpaces is based on the tuple-space work performed by Dr. Gelernter and his team at Yale University during the development of the Linda system. JavaSpaces serve as a "repository" for Java language objects. A new software architecture called an "object flow architecture" can be used with JavaSpaces to create distributed processing systems. It is this type of system that we propose development of to solve the image reformation problem.
Knowledge, skills, and expertise required for this project:

Java programming language V1.3, 1.4
Jini
JavaSpaces
Image file formats (TIF, JPEG, GIF)

For more information on Jini / JavaSpaces technology, the following references are available:

Professional Jini, Sing Li, Wrox Press 2000
Core Jini, W. Keith Edwards, Prentice Hall 2000
JavaSpaces Principles, Patterns, and Practice, Freeman, Hupfer and
Arnold, Prentice Hall, 1999
Equipment Requirements:

If this project is chosen, and development is successful, the
software will be deployed on a dedicated network containing a minimum of
16 MacIntosh G4 cubes running OS X. For development purposes, at
least three machines should be used to demonstrate the
distributed processing paradigm. We recommend the use of Wintel
platforms for development of the actual code. Any combination
of Wintel / Solaris / MacIntosh hardware may be chosen by the
students for demonstration. Lockheed Martin will loan the students some
of these hardware assets for the duration of the project.

Software is available free of charge from Sun Microsystems. Java
1.3, 1.4, Forte for Java IDE and the Jini Starter Kit are all available for
download from the Sun website. Lockheed Martin will provide a suite of
imagery samples to develop and test the system with.

Should you require any further information, please feel free to contact me.
Deliverables:
  1. Design and implement a distributed image reformation system based on Jini/ JavaSpaces technology using an object flow architecture
  2. Support the reformation of byte-stream imagery data into TIF, JPEG and GIF formats.
  3. Demonstrate the viability of the system in a heterogeneous hardware environment consisting of UNIX and Windows based hardware platforms.
  4. Perform timing tests and analysis to demonstrate improved performance over a monolithic imagery reformation algorithm.