Parallel Computing and the IRAS Galaxy Atlas

Citation

Cao, Yu (1997) Parallel Computing and the IRAS Galaxy Atlas. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/nd72-rd29. https://resolver.caltech.edu/CaltechTHESIS:07182025-020527328

Abstract

The Infrared Astronomical Satellite carried out a nearly complete survey of the infrared sky, and the survey data are important for the study of many astrophysical phenomena. However, many data sets at other wavelengths have higher resolutions than that of the co-added IRAS maps (4' - 5'), and high resolution IRAS images are strongly desired both for their own information content and their usefulness in correlation studies.

The HIRES program was developed by the Infrared Processing and Analysis Center (IPAC) to produce high resolution (~ 1') images from IRAS data using the maximum correlation method (MCM). We describe the port of HIRES to the Intel Paragon, a massively parallel supercomputer, and other software tools developed for mass production of HIRES images.

Images produced from the MCM algorithm sometimes suffer from visible striping and ringing artifacts. Correcting detector gain offsets in the reconstruction scheme was found to be effective in suppressing the striping artifacts. A variation of the destriping algorithm was used to subtract zodiacal emission. Using a Burg entropy metric in the image space gave good ringing suppression results for some test cases, but was found to have difficulties with photometry and resolution enhancement and hence not used in subsequent image production. A different ringing suppression algorithm was later developed, which aims to maximize cross log entropy between measured and modeled data. The algorithm suppresses point source ringing, and gave scientifically superior image for the ∝ Ori test field. A partial convergence proof for the log entropy algorithm was achieved.

HIRES images in the 60 and 100 μm wavelength bands were produced for the Galactic plane (-4.7°< b < 4.7°) and the Orion, ρ Ophiuchi, and Taurus-Auriga molecular clouds, using the MCM algorithm plus the gain compensation destriping technique (ringing suppression was not used because the cross log entropy algorithm had not been developed when the processing started). The images and ancillary maps comprise the IRAS Galaxy Atlas. The Atlas images have resolution of approximately 1' - 2'.

Beam matching of the 60 and 100 μm images for the star forming region W3-5 was carried out using the cross-band simulation technique, and a geometrical model was built for a wind-blown shell in W4.

Thesis Files