deriv -- Take the absolute derivative of an image
When comparing an image to a model created by the "blot" task, two primary errors may be expected: the model may be a smeared version of the true image and/or it may be slightly shifted from the true position of the image. The significance of either of these errors depends on the derivative of the image at the point in question (a bright but flat image would be relatively unaffected by either of these two errors).
This task takes the difference of a pixel with each of the four neighbors with which it shares a side and saves the largest absolute value among these differences. The output image is used by the task "driz_cr" to determine the significance of possible blemishes in a data image.
The output image of absolute differences is given the name FILE_deriv, where FILE is the name of the input image.
For an example of the use of this task in reducing dithered WFPC2 observations, see the paper by Fruchter and Mutchler at