PIPT Overview

For effective exploitation of digital imagery data it is essential to process the data using a variety of techniques (typically in an interactive fashion), such as filtering, enhancement, feature extraction, and classification. Thus, there is a great need for a collection of image processing routines which can easily and effectively be used on a variety of data. The Image Processing Toolkit (IPT) was developed in the Image Exploitation group at Rome Laboratory in 1993 in order to meet this need. It provides a basic set of image handling and processing routines and a framework for expanding the capabilities of the toolkit.

While the toolkit provides the necessary functionally and flexibility required for the intended tasks, it is clear that for use in an interactive workstation setting, many important image processing tasks are too slow. This is to be expected, given the large image sizes (commonly up to 10,000 X 10,000) and the complex operations which are required in image exploitation. For example, a simple smoothing operation which removes noise contamination from a 5000 X 5000 pixel image requires over 250 million multiplies and additions, a process which would take several minutes to complete on a common desktop workstation. While this might be acceptable in a batch processing environment, it is an unreasonable time to ask a user to wait in an interactive environment. One cost-effective mechanism for speeding up the computation for such tasks is to utilize existing workstation cluster for parallelism.

The Message-Passing Interface Standard (MPI) is an international effort funded by ARPA and NSF to establish a standard process communication protocol on distributed memory parallel machines. Message passing is a well understood paradigm for parallel processing and a standard interface allows the implementation of portable parallel algorithms. By extending the IP Toolkit using the MPI standard we can assure that implemented routines can be easily ported existing (and new) parallel environments.

The objective of this task is to design, develop and test a parallel/distributed extension to the Image Processing Toolkit (IPT). Due to the use of the MPI standard the Parallel Image Processing Toolkit (PIPT) will remain portable, extensible, and in the public domain.