Space Telescope Science Institute
WFC3 Data Handbook 2.1 May 2011
Table of Contents Previous Next Index Print

WFC3 Data Handbook > Chapter 4: WFC3 Images: Distortion Correction and MultiDrizzle > 4.3 MultiDrizzle Examples

This example, was prepared for the HST Calibration Workshop (Mutchler, M. 2010, HST Calibration Workshop), using data from the HST program 12050 (HH 901), but can be used for any mosaic data sets from the HST Wide Field Camera 3 infrared channel (WFC3/IR). A complete and well-documented set of drizzled output images, prepared by the author for all UVIS and IR filters, are available as High Level Science Products (HLSP) in the Multimission Archive at STScI (MAST):
MultiDrizzle: pipeline versus offline
The MultiDrizzle table (MDRIZTAB) sets parameters for pipeline drizzling, but the user can often produce more optimal results offline by setting parameters tailored to the characteristics of a specific data set, as illustrated here. Data quality flags are generated by steps within the WFC3 calibration pipeline (calwf3), and are stored in the DQ arrays of the flt images used as input for MultiDrizzle. All of the flagged pixels would be excluded during drizzling, but by setting bits (summing any flag types to be included), the user can control which pixels are excluded. For example, the user may set bits=4864 to include pipeline-rejected pixels (flag=4096), saturated pixels (flag=256), and dust motes (flag=512) for an IR data set that did not employ a dither pattern large enough to span them.
The pipeline parameters are generic values set to run on every WFC3 image. Thus the pipeline drz images should be considered quick-look products. Using the basic WFC3 pipeline calibrations (bias, dark, flat), and the “on-the-fly” (OTFR) pipeline-calibrated archival flt images from the HST archive (MAST) as input users can run MultiDrizzle offline. A common reason for reprocessing is to take advantage of subsampled data sets that can be drizzled offline to smaller pixel scales, using a smaller pixel scale and drop fraction (pixfrac), to fully extract all the spatial information contained in them. Similarly, mosaic data sets must be drizzled offline to align and combine images from many different visits (various pointings or different epochs).
Getting started: input images and distortion tables
The sample commands below are shown at a PyRAF prompt (so IRAF commands can also be used), but MultiDrizzle can also be run with python syntax (see the MultiDrizzle Handbook section 5.4). De-archive a set of calibrated flt images using data from the HST program 12050 (HH 901): typically using the same instrument modes and filter, with exposure times within 20% of each other. Make an input list in your working directory:
Download the latest distortion reference table, and specify it’s path name in your image headers (IDCTAB keyword). You may wish to make your own local iref directory and add reference files there:
pyraf> hedit i*flt.fts[0] IDCTAB "iref$t20100519_ir_idc.fits"
MultiDrizzle and the core drizzle task (Fruchter & Hook, 2002) are available in the STSDAS dither package. Load multidrizzle and set parameters as suggested below. Parameters can be saved in a “par” file to record and reuse them later, or a command script can be used to quickly make parameter changes and iteratively reprocess images. The initial run produces the single sci images needed to measure shifts.
Image registration and shift file
MultiDrizzle uses the World Coordinate System (WCS) information in each image header to align the images. However, any data set including images taken in different visits or epochs will have small misalignments between visits, due to errors in the cataloged positions of the different guide stars used. Objects in the overlaps between different pointings can be used to measure “delta” shifts and rotations in each image or update the World Coordinate System in each image. These shifts and rotations can then be used to register the images for drizzle combination. Initial shifts can be quickly measured by visually selecting a few ideal objects (e.g. stars which are unsaturated and uncontaminated by cosmic rays) in the undistorted single_sci images produced by the driz_separate step. The results must be put into a shiftfile with the following format:
After applying the initial x,y shifts (in the 2nd and 3rd columns above), tweakshifts or geomap can be run on the single_sci images to further refine the shifts, and also solve for small rotational or scale offsets (the 3rd and 4th columns in the shift file).
Key drizzling parameters
The following are a key subset of MultiDrizzle parameters optimized for an IR mosaic dataset, with output at both the native detector pixel scale, and an enhanced scale:
Drizzle to native scale, and make a scaled sum image (Figure 4.3) for diagnostic purposes:
Although this observing program did not employ an optimally subsampling dither pattern, much of the nebula resides in the overlap areas with extra sampling, so the following is an example of drizzling to enhance resolution. The smaller pixel scale is 62% of the detector pixel scale, and happens to be conveniently twice the UVIS scale. Also note the smaller pixfrac “drop size” and alternate kernel:
pyraf> multidrizzle input=’@list_flt_f128n’ output=’hh901x_wfc3_f128n’ final_scale=0.07920 final_pixfrac=0.8 final_kernel=’gaussian’
Inspecting output and iteration
To visually verify the quality of your drizzled output, blink your drz_sci, drz_weight, and drz_sum images (Figure 4.3) to look for signs of bad rejections (too much or too little), misregistration, or other signs that some parameters may need to be adjusted. The median image produced by MultiDrizzle is also worth inspecting: it is used to identify cosmic rays and other bad pixels to be rejected, so it should look almost as good as the final drizzled image. The rms of the exposure weight map (drz weight) should typically be under 30% of the mean, to ensure photometric consistency throughout the image. Several trial and error iterations of are typically required to arrive at optimal results.
Figure 4.3: F128N sum image, which shows the relationships between IR detector features, pointing overlaps, and the target. Blinking this image against the clean drizzled output and the corresponding exposure weight map provides a good initial diagnostic inspection.
One quadrant of the mosaic is rotated with respect to the others: that pointing failed initially, and had to be repeated several weeks later with a different nominal roll angle.
More examples on how to use MultiDrizzle with WFC3 data will be made available on the following Web site:
Meanwhile, users may also find the examples included in Chapter 6 of the MultiDrizzle Handbook (Fruchter & Sosey et al. 2009) useful. The examples developed for the ACS CCDs in Section 6.2 may be used as a guide for reprocessing images from the WFC3 UVIS detector. Example 1 describes the steps required to improve the final drizzle sampling, Example 2 describes how to refine the image alignment (especially important for images obtained in separate visits), and Example 3 describes how to fine-tune the cosmic-ray flagging and create an image mosaic. For WFC3-IR images, the NICMOS example in Section 6.3 may be used as a guide for reprocessing.

Table of Contents Previous Next Index Print