Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.sao.ru/precise/Midas_doc/doc/94NOV/vol2/node214.html
Дата изменения: Fri Feb 23 14:02:01 1996 Дата индексирования: Tue Oct 2 18:36:47 2012 Кодировка: Поисковые слова: п п п п п п п п п п п п п п п п п п п п п п п п п п п п п п п п п п п п п п |
Modern solid-state detectors produce data which differ in two main aspects from classical photographic images: (a) a much higher S/N (signal to noise ratio) and (b) the rigid spatial digitisation of the incoming flux. The importance of (b) becomes evident if one considers that for most astronomical instrument/detector combinations the sampling of the point--spread function is hardly adequate and usually just fulfills the Nyquist criterion which states that there should be at least two detector elements to a resolution element. Thus, in good data, the information content of each pixel is highly significant, and one must ensure that this information is not lost or degraded during the data reduction process.
While the user wants to be assured of this, it appears unrealistic to expect that complicated algorithms can always be documented to the sub--pixel level. With the accelerating exchange and export of software in astronomy, inquiries to the people who have written or are maintaining the software are often difficult or impractical. The only solution is testing. This must be done on some suitable test images so that the results can immediately be checked with, e.g., a pocket calculator. This evidently requires images with some simple geometrical patterns.
The primary use of such images is in assessing accuracy of operations involving transformations of images.
The generation of such test images is a non--trivial task, both on the
conceptualisation and on the implementation levels because the test
images must be independently conceived of the software items which
will be subsequently used on them, and the numerical precision of the
test images (we have aimed at 1 in 10) must exceed the S/N of
typical observations by an order of magnitude. We therefore offer a
small set of programs for the generation of such test images and, in
cases where the required execution time is prohibitive, the frames
themselves. The objective of the package is to allow tests of all
routines which do not maintain the identity of a pixel such as
rebinning, rotation and filtering. Fitting can also be considered. The
usage of the programs and images that we provide as a reference test
package should also facilitate documentation of possible application
software malfunctions and their subsequent repair. We provide in detail below
the images constructed, with the precisions obtained, and mention
implementation details where relevant.
It is clearly not possible to provide every image configuration of interest but it may be noted that the images which are provided may be combined in any desired manner. This refers in particular to the rotating or flipping of the patterns described below, or the adding of various backgrounds (also provided) to these images. Images with spatial frequency spectra which come close to those of real data can be obtained by filtering.
The commands described in the following sections are activated by setting the GEOTEST context, by using the command SET/CONTEXT GEOTEST in a MIDAS session.