wfs_combine step combines a pair of dithered Wavefront Sensing and Control (WFS&C) images.
The input images are aligned with one another and then combined using a pixel
replacement technique, described in detail below. The images are aligned to only the nearest
integer pixel in each direction. No sub-pixel resampling is done.
Due to the WFS dither patterns oscillating between two locations, the first image of the pair will oscillate between the two dither locations. Because the WSS software works in pixel space, we need to change which input image is “image 1” to get star to have the same pixel location in the output image. When the input parameter “flip_dithers” is set to True (the default) and the x offset between image 1 and image 2 is negative, the two images will be switched before any processing is performed.
Creation of the output combined image is a three-step process: first the offsets between the images are computed using the World Coordinate System, then the offsets are used to shift image 2 to be in alignment with image 1, and finally the aligned data from the two images are combined.
The WCS transforms of each image are used to compute the RA/Dec values for the center pixel in image 1, and then the pixel indexes of those RA/Dec values are computed in image 2. The difference in the pixel indexes, rounded to the nearest whole pixel, is used as the nominal offsets in the x/y image axes.
If the optional argument “–do_refine” is set to
True, the nominal offsets are empirically
refined using a cross-correlation technique. The steps in the refinement are as follows:
Create a smoothed version of image 1 using a Gaussian kernel.
Find the approximate centroid of the source in image 1 by computing the mean pixel coordinates, separately in the x and y axes, of all pixel values that are above 50% of the peak signal in the smoothed image.
Create subarrays from image 1 and 2 centered on the computed source centroid.
Create the cross-correlation image of the two input images.
- Find the peak intensity of the cross-correlation image and use this to determine the
- Use the find the difference between the cross-correlation pixel offsets and the WCS offsets.
Add these deltas to the nominal offsets computed from the WCS info to form the refined offsets.
Creating the Combined Image¶
The image 2 data are shifted using the pixel offsets computed above, in order to align it with image 1. For each pixel location in image 1, the output combined image is populated using the following logic:
If the pixel values in both image 1 and 2 are good, i.e. DQ=0, the output SCI and ERR image values are the average of the input SCI and ERR values, respectively, and the output DQ is set to 0.
If the image 1 pixel is bad (DQ>0) and the image 2 pixel is good, the output SCI and ERR image values are copied from image 2, and the output DQ is set to 0.
If the image 1 pixel is good (DQ=0) and the image 2 pixel is bad, the output SCI and ERR image values are copied from image 1, and the output DQ is set to 0.
If both image 1 and 2 pixels are bad (DQ>0), the output SCI and ERR image values are set to 0 and the output DQ contains the combination of input DQ values, as well as the “DO_NOT_USE” flag.
Upon successful completion of this step, the status keyword S_WFSCOM will be set to “COMPLETE” in the output image header.
2D calibrated images¶
- Data model
- File suffix
The input to
wfs_combine is a pair of calibrated (“_cal”) exposures, specified
via an ASN file. The ASN file may contain a list of several combined products to be created, in
which case the step will loop over each set of inputs, creating a combined output for each pair.
2D combined image¶
- Data model
- File suffix
The output is the combined image, using the product type suffix “_wfscmb.”