Objective: Our goal is to remove cirrus emission from images of the sky generated by the Infrared Astronomy Satellite (IRAS). The cirrus emission looks remarkably like the cirrus clouds which form on Earth, but is caused by cold dust grains in our Milky Way galaxy. This infrared cirrus emission obscures our view of the universe beyond the Milky Way, and by removing it we will create a valuable new public archive, and we may even reveal new, unusual infrared objects.
Approach: Previous attempts to remove the cirrus emission have failed because the emission is present on all angular scales. Our approach is to apply the techniques of morphological image processing (a.k.a. mathematical morphology). Morphological image processing is a relatively new set of tools for analyzing form and structure in images. The techniques can be computationally intensive, and so we are implementing the morphology tools on the HPCC ESS testbeds, in particular the MasPar MP-2.
Accomplishments: We have dramatically improved our prototype morphological cirrus filter. This improvement was largely enabled by the tremendously faster performance of the MasPar compared to an earlier workstation implementation. We have filtered a few dozen IRAS images and are now analyzing the nature of the objects we find. This analysis involves comparing our source positions with large catalogs which are available via the Internet. We are finding many galaxies which were previously discovered at optical wavelengths, but which were previously very obscured in the infrared by the cirrus. Preliminary analytical testing shows that the filter is able to recover obscured galaxies with an accuracy of better than a few percent.
A paper describing a detailed comparison of different MasPar implementations of morphological filtering was submitted for review to the Frontiers of Massively Parallel Computation meeting to be held in February, 1995. Presentations were made to an American Astronomical Society meeting in May, 1994 and to Astronomical Data Analysis Software and Systems symposia in October, 1993 and September, 1994. The morphology kernels were selected to be part of the ESS Parallel Benchmark Suite, and are being benchmarked on a variety of platforms.
Significance: We hope to improve our knowledge of the infrared brightnesses of galaxies, add to our understanding of the cirrus emission, and possibly even discover new astronomical objects. We will also publicly deliver the morphology kernel routines optimized for a variety of HPC platforms.
Status/Plans: We are continuing analytical testing to determine the accuracy and reliability of our filter. We expect to perform production filtering of the entire IRAS database at one and perhaps two far infrared wavelengths. This new astronomical archive will be made publicly available.
Point of Contact:
Return to the PREVIOUS PAGE