Документ взят из кэша поисковой машины. Адрес оригинального документа : http://xmm.vilspa.esa.es/sas/5.4.1/doc/asmooth/node8.html
Дата изменения: Fri Jan 10 22:48:37 2003
Дата индексирования: Sat Dec 22 08:11:31 2007
Кодировка:

XMM-Newton Science Analysis System


asmooth (asmooth-2.10.2) [xmmsas_20030110_1802-5.4.1]

Meta Index / Home Page / Description / Adaptive smoothing

Smoothing of mosaics and other images with exposure steps

It is now possible to handle these by supplying asmooth with an exposure map. The task divides the input image by this map and smooths the result, thus avoiding `smearing out' steps in brightness due to discontinuities in the exposure. So, one might do for example

emosaic imagesets='image_1.ds image_2.ds' mosaicedset=image_mosaic.ds

emosaic imagesets='expmap_1.ds expmap_2.ds' mosaicedset=expmap_mosaic.ds

asmooth inset=image_mosaic.ds withexpimageset=yes expimageset=expmap_mosaic.ds

There are in principle two obvious ways to treat the variance image (regardless of whether this has been supplied by the user or generated internally). If expmapuse is set to `samesnr', the variance image is divided by the square of the exposure map. The effect of this is to leave the output image with the same signal-to-noise ratio, just as in the case where no exposure map is supplied. The amplitude of the remaining noise in the smoothed image will look everywhere about the same, but the size of these noise variations will look smaller in those parts of the image with longer exposure. The second choice is to set expmapuse to `samesize'. In this case, the initial variance image is divided by the exposure map and then divided again by the maximum value of the exposure map. The effect of this is to produce noise which has everywhere about the same spatial size, but with an amplitude which is less in those parts of the image with longer exposure. The user must decide which `look' they prefer.

A remaining irritation with this procedure is that XMM images include a non-vignetted background component (due to X-ray fluorescence of the substrate, among other things). Dividing this by the exposure map of the image (which tends to be dominated by the hump of the vignetting function) leaves a `hollow' in the centre. Ideally one would wish to subtract some approximately constant value from each image before submitting it to asmooth. A background map may be obtained by running eexpmap with withvignetting='no', then multiplying the result by an estimate of the background rate per image pixel. Unfortunately this must all be done `by hand' at the moment since we don't yet have good means of calculating the relative contributions of vignetted vs. non-vignetted background.