Astrometric
|
MultiDrizzle introduces astrometric errors
on the order of a few pixels (typically from 1-3 pixels) when creating an output drizzled image with a non-default orientation. Behavior: The default orientation for the output product gets determined from the distortion corrected orientation of the first input image. However, the user can specify the orientation of the final product by setting the 'final_rot' parameter to a non-blank value. It has been found that when 'final_rot' gets used, the output image will have an offset introduced such that the RA and Dec positions of sources will no longer be consistent with the original distortion corrected input positions. Workaround: None. It can be avoided by using the default output image. |
Cosmic-Ray
|
The identification of cosmic-rays during
MultiDrizzle processing can be skewed slightly high sky values in the image. Behavior: The algorithms used for identifying bad pixels and cosmic-rays in the input images relies on comparing the median image with the original distortion blotted back in to the original distorted input images. This comparison in MultiDrizzle, though, occurs after subtractingthe sky the input image, yet the determination of the outliers from statistical noise models took into account the sky values which have already been removed. This can result in pixels being flagged as 'bad' by MultiDrizzle when they may actually fine; for example, they may simply be on the edge of a PSF feature. Workaround: None. |
Binned Data
|
The distortion model will not be applied
correctly to all binned data; in particular, WFC3 data. Behavior: It was found during testing in preparation for pipeline use of MultiDrizzle for WFC3 data, that binned WFC3/UVIS images using 2x2 binning were not being mosaiced correctly, introducing a gap between the WFC3/UVIS chips of nearly 2000 pixels. This turned out to be a bug in the way the distortion model gets applied to binned data. WFC3 data taken using 3x3 binning, though, has data which appears to have the chips placed correctly in an image which stretches 3 times the height of the corrected chips. Workaround: None. |
Subarray data
|
The distortion model does not get properly interpreted
for use with subarray observations taken by any instrument including WFC3 and ACS. Behavior: The distortion model for subarray observations (WFC3, ACS, and even STIS data) does not get applied relative to the image's reference position as needed. This causes a mis-application of the distortion model such that corrected subarray data does not align with a full-frame image of the same field of view. It has been seen to induce a scale difference of nearly 3%, and a rotation of nearly a full degree relative to a properly distortion-corrected direct image of the same field of view. Workaround: None |
|
The distortion model does not get properly interpreted
for use with ACS polarizer observations. Behavior: The distortion model for ACS polarizer observations does not get correctly applied as the code always interprets the distortion model relative to a full-frame image while the polarizer distortion model was only specified for the subarray mode used for polarizer data. This causes a mis-application of the distortion model such that corrected polarizer image does not align with a full-frame image of the same field of view. It has been seen to induce a scale difference of nearly 3%, and can also induce a rotation of up to a full degree relative to a properly distortion-corrected direct image of the same field of view. Workaround: None |
MinMed masking
|
MultiDrizzle implements a the 'minmed' algorithm
for creating the median image, which combines a minimum image with
a clipped median. Behavior: By default, the highest input pixel gets clipped by this method, resulting in an output where regions with only input from 1 image gets masked out, instead of being retained as the minimum value. Thus, only regions where more than 1 image overlaps gets retained in the output median image when using 'minmed'. Workaround: None. |
Missing Pixels |
MultiDrizzle computes
the parameters necessary for running 'drizzle' using PyDrizzle.
The user can also specify the RA/Dec position of the output frame. Behavior: Some pixels get reported as 'dropped' during the single and final drizzle steps. If a user specifies the RA/Dec, MultiDrizzle will shift the images accordingly. This situation has been improved significantly, but it still can underestimate the final output size for this situation resulting in dropped pixels/rows. Workaround: A new output size can be specified which will be large enough to accomodate the entire output frame, if significant data is lost. |
Zero Exposure Time Images
|
MultiDrizzle performs all image combination in
units of electrons per second, requiring (in most cases) conversion
using the exposure time. However, some associations include images
where the exposure time gets reported as 0.0 seconds. Behavior: All input images with zero exposure time are noted, reported to the user, then ignored for all MultiDrizzle processing. Workaround: The user can verify whether there is any valid data in the zero-exposure time input image. If the image contains useful data, the EXPTIME keyword in each SCI extension of that input image can be updated with a non-zero value that best approximates the actual exposure time used. |
Output Units for
|
MultiDrizzle performs
all image combinations in units of electrons per second, units which
are not used for IR data. The conversion of NICMOS and WFC3 IR
data to electrons per second within MultiDrizzle required information
about the units of the input images. The final product, and
only the final product, would then be converted back to the set of units
which the user specified in the 'final_units' and 'proc_units' parameters.
Behavior: MultiDrizzle Version 3.3.5 released in Nov 2009 properly handled WFC3 IR data, but did not correctly scale the NICMOS data for the final output. This problem was resolved for the next public release of MultiDrizzle which should be available in the Summer of 2010. Workaround: None |
Questions? Contact help@stsci.edu
26 March 2010