STScI Logo

crossdriz stsdas.analysis.dither



crossdriz -- Builds a set of cross-correlation images from one single-group image and one single-group reference image, and an optional set of rotation angles.


crossdriz image refimage rootname


image [file list/template]
File list/template with input image names. Only single-group images are supported.
refimage [file name]
Input single-group reference image.
rootname [file list]
File list with the base names for the output cross-correlation images. Each base name is paired with the corresponding image in the input list, and generates a family of cross-correlation images, one for each rotation angle. Each of these images will have its name built by appending "_NN" to the corresponding base name, where NN is a two digit sequential counter (01, 02, 03,...). If only one rotation angle is specified, no suffixes are appended to the base names.
(dinp = yes) [boolean]
Drizzle input image(s) ? When set to "yes", the task will take care of drizzling all input images but only if needed, that is, only if geometrical correction / resizing / rotation is taking place. If set to "no", drizzling of input images is disabled and the task will issue a warning if geometrical correction / resizing / rotation operations are also being asked for, but will proceed anyway.
(dref = yes) [boolean]
Drizzle reference image ? When set to "yes", the task will take care of drizzling the reference image only if needed, that is, only if geometrical correction and/or resizing is taking place. If set to "no", drizzling of the reference image is disabled and the task will issue a warning if geometrical correction and/or resizing operations are also being asked for, but will proceed anyway.
(margin = 50) [int, min=0]
To avoid using the image edges, which in WFPC2 images are often quite noisy, the cross-correlation can be computed from a subsection of the input and reference images. The task uses a subsection which is the original image stripped down of this margin size. The cross-correlation images will have their size smaller than the input images, by the same amount.
(tapersz = 50) [int, min=1]
Size of cosine bell tapering window used in the (already sectioned) image.
(mintheta = 0.) [real]
Starting rotation angle for the cross-correlation image sequence (in degrees).
(maxtheta = 0.) [real]
Ending rotation angle. If set to the same value in mintheta, only one cross-correlation, corresponding to that specific rotation angle, will be generated.
(stptheta = 0.1) [real]
Rotation angle step.
(coeffs = "header") [string]
File with geometrical correction coefficients. If set to "header", the Trauger coefficients for the proper WFPC chip will be used. An empty string ("") implies that no geometrical correction will take place.
(lambda = 555.0) [real]
Effective wavelength (nm). Used only by Trauger coefficients.
(xout = INDEF, yout = INDEF) [real]
Drizzled image size. If set to INDEF, the original size is kept.
(pad = no) [boolean]
Pad images with zeros to prevent wraparound effects ? If selected, this will increase the amount of memory and execution time use by the task.
(tempdir = "tmp$") [path name]
Path to directory where temporary images and tables are kept.
(verbose = yes) [boolean]
Print log of what is beign done ?


The position of the peak of the cross-correlation of two images directly provides the relative shift between the two images. However, due to the geometric distortion of the WFPC2 cameras, and possible rotation between images, simple cross-correlation frequently will not provide an optimal (or even correct) estimate of the offset between two HST images. This task uses the drizzle program to remove the geometric distortion of the images, and to rotate the input image to various test rotation angles, if the user suspects a difference in orientation between the input and reference images. This task also takes care of the image preparation necessary to minimize edge effects in the final cross-correlation image.

The steps involved are:

Drizzle the reference image.
Extract a section of it (to get rid of corrupted edges) and taper the result using a cosine bell function (to reduce edge effects).
For each selected rotation angle, repeat the steps above for the input image, and cross-correlate the result with the processed reference image, resulting in a cross-correlation "image".

The resulting cross-correlation images have the test rotation angle added to their header using the keyword "ROTANGLE". This keyword is used by task "rotfind".

The task can process image lists as well as single images. The input image list and output cross-correlation image rootname list must be paired one-to-one in this case. Only one reference image will be used, thus all cross-correlations will refer to this single image.

Crossdriz only works on single group images (though the drizzle task will automatically use the correct geometric distortion in cases where the image has been extracted from a multi-group WFPC2 image). In order to cross-correlate multi-group WFPC2 images users should employ the offset task, which calls crossdriz. Offset, however, does not perform rotation.

In cases where the relative shift between images is only a few pixels and no rotation is suspected between images, users may wish to turn off the drizzle option in order to significantly speed-up the running of crossdriz.


di> crossdriz image2,image3,image4 image1 im2x1,im3x1,im4x1 dinp=no dref=no

cross-correlates the images in the first list with image1, and puts the output in the last list. The images would not be drizzled before cross-correlation.

di> crossdriz image2 image1 im1x2 minthet=1.4 maxthet=1.6 stpthet=0.05

The above would produce five cross-correlation images, corresponding to test rotation angles of 1.4, 1.45, 1.5, 1.55, and 1.6 degrees. The output images could be given to rotfind for determination of the best-fit rotation angle.


The drizzle task parameters are temporarily modified by this task, and are reset to the original values at the end of the task. If crossdriz crashes, this last step may not be performed, and the drizzle parameters may left modified.


This task was written by I.Busko in collaboration with Andy Fruchter. More information can be found in A. S. Fruchter, R. N. Hook, I. C. Busko, and M. Mutchler, 1997, "A Package for the Reduction of Dithered Undersampled Images", in "The 1997 HST Calibration Workshop", S. Casertano et al., ed. (Baltimore: Space Telescope Science Institute), in press. See



Source Code · Package Help · Search Form · STSDAS