Документ взят из кэша поисковой машины. Адрес оригинального документа : http://xmm.vilspa.esa.es/sas/5.4.1/doc/asmooth/node9.html
Дата изменения: Fri Jan 10 22:48:37 2003
Дата индексирования: Sat Dec 22 08:24:02 2007
Кодировка:
Поисковые слова: http astrokuban.info astrokuban

XMM-Newton Science Analysis System


asmooth (asmooth-2.10.2) [xmmsas_20030110_1802-5.4.1]

Meta Index / Home Page / Description / Adaptive smoothing

Masking

Images often include large `black' areas outside the CCD chip boundaries. If these are not masked off, the smoothed image will extend out into these non-CCD regions. The task therefore allows the user to load a detector or `output' mask via the parameters withoutmaskset and outmaskset. The output smoothed image is masked by this array such that zeros are written where the mask is false.

A related effect is the 'drooping' of the values of a smoothed image near to steps in the input. Where the steps are between two non-zero levels of exposure, this smearing of steps can be removed by dividing the image by an exposure map prior to smoothing (see section 3.1.4). However this will obviously not work if the exposure step is from some non-zero value directly to zero. The best way to cope with such steps is to divide the smoothed image by a similarly-smoothed mask whose TRUE values are restricted to pixels where the image is $> 0$. This `input' mask can be supplied to the task via the parameters withinmaskset and inmaskset. This technique has the additional useful property of `papering over' small holes in the image. To achieve this, such pixels should be included in the output mask but not the input mask. Thus if it is desired to smoothly interpolate over holes in the image due, for example, to the removal of sources, or gaps between CCDs, then the output mask should be TRUE at these places, but the input mask should be FALSE.

Note however that, if the discrepancy between inmaskset and outmaskset is too great, the value of the weighting function may become very small far from the edges of holes in inmaskset. This can lead to numerical imprecision in the smoothing. The task monitors this process and, if it seems likely that the weight will be too small for the desired precision (and provided the user has selected inmaskset=`adaptive' so as to allow the variation of sigma), asmooth will increase the sigma value of the smoothing kernel until the precision is achieved. If this is not possible, a zero is written to the output at that position and the pixel is set to TRUE in a map which is obtainable via the writebadmaskset and writebadmaskset parameters.

The mask images inmaskset and inmaskset can have any of the data types of the input image, in fact it is recommended that the exposure map be used as the output mask. The task creates a logical mask internally by setting to false those pixels for which the mask $<=0$.

If an input mask is not supplied, it is set internally to be the same as the output mask.