Previous abstract Next abstract

Session 55 - Space Instrumentation.
Display session, Wednesday, June 12
Tripp Commons,

[55.14] ``Cirrus'' Removal from IRAS Images Using Mathematical Morphology and Wavelet Analysis

L. X. He, P. N. Appleton, J. P. Basart (ISU), J. A. Pedelty (NASA/GSFC)

The presence of diffuse far-infrared "cirrus" emission in IRAS survey data has hampered the identification and accurate determination of fluxes for background extra-galactic sources, especially near the Galactic Plane. We present a self-consistent filtering process for the removal of "cirrus" emission consisting of size information extraction and classification. The process removes the flux corresponding to the average "cirrus" component at each pixel in the IRAS survey data at 60 and 100 microns. Size information is extracted using either a morphological sieving process or by wavelet analysis. The mathematical morphological method uses structuring elements of Gaussian, cylindric, or hemispheric shape. The wavelet method, however, extracts localized information from both the spatial frequency and the spatial domains. Classification of the size information is carried out by a k-means method. Using the classification scheme, we are able to characterize the size information of typical Galactic "cirrus" and remove it from the survey images. The Mathematical Morphology approach was found to be the most successful technique. In order to test the filter, we have superimposed "model" distributions of extragalactic emission onto images with markedly different degrees of "cirrus" contamination (source templates included extended extra-galactic sources such as the emission profile of the galaxy M101). These tests show that if a sigmoidal weighting function is used, intensity information for background sources is retained to better than 15%, even in regions of strong, rapidly varying "cirrus" emission. A complete set of filtered images for the 60 and 100 micron bands of the IRAS Survey (including the Galactic Plane) will be shortly available on CDROM by the authors. This work was funded by NASA-HPCC grant NAG-5-2286.

Program listing for Wednesday