Physics
Scientific paper
Dec 2009
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2009agufm.p23a1231c&link_type=abstract
American Geophysical Union, Fall Meeting 2009, abstract #P23A-1231
Physics
[1640] Global Change / Remote Sensing, [6225] Planetary Sciences: Solar System Objects / Mars, [6297] Planetary Sciences: Solar System Objects / Instruments And Techniques
Scientific paper
The Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) is affected by a common artifact in "push-broom" sensors, the so-called "spectral smile". As a consequence, both central wavelength and spectral width of the spectral response vary along the across-track dimension, thus giving rise to a shifting and smoothing of spectra (see Fig. 1 (left)). In fact, both effects are greater for spectra on the edges, while they are minimum for data acquired by central detectors, the so-called "sweet spot". The prior artifacts become particularly critical for Martian observations which contain steep spectra such as CO2 ice-rich polar images. Fig. 1 (right) shows the horizontal brightness gradient which appears in every band corresponding to a steep portion of spectra. The correction of CRISM spectral smile is addressed using a two-step method which aims at modifying data sensibly in order to mimic the optimal CRISM response. First, all spectra, which are previously interpolated by cubic splines, are resampled to the "sweet spot" wavelengths in order to overcome the spectra shift. Secondly, the non-uniform spectral width is overcome by mimicking an increase of spectral resolution thanks to a spectral sharpening. In order to minimize noise, only bands particularly suffering from smile are selected. First, bands corresponding to the outliers of the Minimum Noise Transformation (MNF) eigenvector, which corresponds to the MNF band related to smile (MNF-smile), are selected. Then, a spectral neighborhood Θi, which takes into account the local spectral convexity or concavity, is defined for every selected band in order to maximize spectral shape preservation. The proposed sharpening technique takes into account both the instrument parameters and the observed spectra. First, every reflectance value belonging to a Θi is reevaluated by a sharpening which depends on a ratio of the spectral width of the current detector and the "sweet spot" one. Then, the optimal degree of sharpening for every Θi is determined thanks to a loop of sharpening procedures, which is assessed by the examination of an estimation of the smile energy (the MNF-smile eigenvalue). As a matter of fact, a higher sharpening is performed on Θi as long as the smile energy decreases. Experiments on CRISM data show remarkable results regarding the decrease of smile energy (up to 80%) and the spectral shape preservation. In fact, initial smile-affected spectra do no longer show shifting nor smoothing (see Fig. 2). Line-averaged spectra and band 155 of FRT5AE3_07 showing spectral smile effects
Line-averaged spectra and band 155 of smile-corrected FRT5AE3_07
Ceamanos X.
Doute Sylvain
No associations
LandOfFree
SPECTRAL SMILE CORRECTION IN CRISM HYPERSPECTRAL IMAGES does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with SPECTRAL SMILE CORRECTION IN CRISM HYPERSPECTRAL IMAGES, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and SPECTRAL SMILE CORRECTION IN CRISM HYPERSPECTRAL IMAGES will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1768968