Cosmic Rays

The package provides three major methods of cosmic ray identification:

• The first approach is to apply two median filters to the data frame. One in the X direction (along rows) and the other in the Y direction (along columns). Both of these median images are then divided into the original and the resulting image histogrammed. A threshold point on the displayed histogram is then selected. All pixels generating samples above the clip threshold are flagged as cosmic rays. This method does not rely on the echelle nature of the image and may be used on data frames of non-spectral type with some success.

• The second cosmic ray removal routine, is more sophisticated in its approach and requires that the order traces are already known. It works by examining the variations from expected intensity (calculated from the average profile) and then calculating the probability that such variations are due to chance.

  Significant pixels are then rejected and the whole order is re-processed. Iterations continue until the overall significance of all outliers in the order is below a user-supplied threshold. Each order is processed independently in two parts, first the sky pixels, and then the object pixels.

  The significance of a fit is calculated by assuming a Gaussian distribution of deviations from expected intensities and using the Kolmogorov-Smirnov statistic.

  A post processor is used to counteract the tendency of cosmic ray identifiers to be fooled by sky line pixels. This task examines the geometry of all 'connected' cosmic ray pixels and restores any which it adjudges are actually due to bright sky emission lines.

• If you have many `identical' object frames then these may be processed all together to provide a more reliable cosmic ray rejection. The command

        $ECHOMOP_SOURCE/decos_many
  

  provides this facility and should be the preferred method when multiple frames are available. The procedure calculates a median image and then rejects pixels which deviate by more than n sigma from the median value (note that this means that all the frames must be of equal exposure time). Bad pixels are flagged by placing values in the quality array of the frame concerned.