Comparison of de-noising techniques for FIRST images
Data obtained through scientific observations are often contaminated by noise and artifacts from various sources. As a result, a first step in mining these data is to isolate the signal of interest by minimizing the effects of the contaminations. Once the data has been cleaned or de-noised, data mining can proceed as usual. In this paper, we describe our work in denoising astronomical images from the Faint Images of the Radio Sky at Twenty-Centimeters (FIRST) survey. We are mining this survey to detect radio-emitting galaxies with a bent-double morphology. This task is made difficult by the noise in the images caused by the processing of the sensor data. We compare three different approaches to de-noising: thresholding of wavelet coefficients advocated in the statistical community, traditional Altering methods used in the image processing community, and a simple thresholding scheme proposed by FIRST astronomers. While each approach has its merits and pitfalls, we found that for our purpose, the simple thresholding scheme worked relatively well for the FIRST dataset.
- Research Organization:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- US Department of Energy (US)
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 15006878
- Report Number(s):
- UCRL-JC-142085; TRN: US200412%%275
- Resource Relation:
- Conference: Third Workshop on Mining Specific Datasets, within the First SIAM International Conference on Data Mining, Chicago, IL (US), 04/07/2001; Other Information: PBD: 22 Jan 2001
- Country of Publication:
- United States
- Language:
- English
Similar Records
The angular two-point correlation function for the first radio survey
On the design and implementation of a parallel, object-oriented, image processing toolkit