The main processing stages of the Mosaic pipeline are as follows:
Interpolation
The input images are interpolated onto the output grid defined by the FIF table. This interpolation is performed by the Mosaic Interpolate module, and there are four interpolation schemes available. This module also corrects for geometric distortion.
Outlier Detection
MOPEX employs four outlier detection methods to detect radiation hits, bad pixels and moving objects. Any number of these methods can be run on the data. The rejected pixels are flagged in the RMask, and the user can select which of the rejection schemes to set as "fatal".
Reinterpolation
The interpolated images are updated to include the outlier detection information. Only the pixels affected by the outlier rejection are recomputed.
Coaddition
The interpolated images are co-added to create one mosaic image. There are several co-addition schemes available, but the default is a straight average of the input pixel values. The interpolated uncertainty images are also co-added into one uncertainty mosaic, and the coverage maps are co-added to produce a coverage mosaic.
Combining
In cases with large mosaics, the coaddition step must be carried out in "tiles" to avoid running out of memory. This final step combines the tiles into one final mosaic. Note that even if the coaddition has been carried out in a single tile, this step must still be run in order to produce the expected output.
