The PROS system provides the capability to keep track of the uncertainties in image files through error arrays. For convenience, the squares of the errors are actually stored in these arrays.
There is a task, errcreate, that will estimate the errors in an image and produce an error array. Two estimates are available, selected by a parameter in the image header. The default estimate assumes that the counts in the image result from a Poisson process; this estimate is the most reliable one when there are few counts per pixel. The other estimate assumes that the counts result from a Gaussian process. See ``help explain_errors'' for details of the two algorithms and an example of how to change the default error calculation from Poisson to Gaussian.
Some tasks that transform images, such as imsmooth, imcompress, and imreplicate have the option to produce a resultant error array, derived from the original image's error array. Note, however, that the imcalc task does not produce an error array. The user is responsible for making one that reflects the operations that have been performed on the image. Some analysis tasks such as imcnts and imcontour, can use error arrays as part of their input.
See section on errcreate for further discussion, and read the help files ``explain_errors'' and ``errcreate''.