Experimental Approach to the Selection of the Components in the Minimum Noise Fraction

• Published in: IEEE transactions on geoscience and remote sensing Vol. 47; no. 1; pp. 153 - 160
• Main Authors: Amato, U., Cavalli, R.M., Palombo, A., Pignatti, S., Santini, F.
• Format: Journal Article
• Language: English
• Published: New York, NY IEEE 01-01-2009; Institute of Electrical and Electronics Engineers; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Aircraft; Aircraft components; Algorithms; Applied geophysics; Atmospheric measurements; Atmospherics; Earth sciences; Earth, ocean, space; Estimating; Exact sciences and technology; Image enhancement; image processing; image restoration; Image sensors; Imaging spectrometers; Infrared imaging; Infrared spectra; Internal geophysics; Libraries; Noise; Noise measurement; Optical imaging; Optimization; remote sensing; Satellites; Spectroscopy; Studies; image processing; noise; remote sensing; image restoration; Image enhancement; algorithms; coupling; spectra; image enhancement; Space remote sensing; imagery; libraries; variability
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2008.2002953

An experimental method to select the number of principal components in minimum noise fraction (MNF) is proposed to process images measured by imagery sensors onboard aircraft or satellites. The method is based on an experimental measurement by spectrometers in dark conditions from which noise structure can be estimated. To represent typical land conditions and atmospheric variability, a significative data set of synthetic noise-free images based on real Multispectral Infrared and Visible Imaging Spectrometer images is built. To this purpose, a subset of spectra is selected within some public libraries that well represent the simulated images. By coupling these synthetic images and estimated noise, the optimal number of components in MNF can be obtained. In order to have an objective (fully data driven) procedure, some criteria are proposed, and the results are validated to estimate the number of components without relying on ancillary data. The whole procedure is made computationally feasible by some simplifications that are introduced. A comparison with a state-of-the-art algorithm for estimating the optimal number of components is also made.