Space Telescope Science Institute
DrizzlePac 2012 Handbook
Table of Contents Previous Next Index Print

The DrizzlePac Handbook > Chapter 7: Examples: Using DrizzlePac for Combining Images > 7.4 ACS/WFC: Optimizing the Image Sampling for a Single Visit

This example describes the combination of four ACS/WFC images, obtained using the ACS-WFC-DITHER-LINE pattern, which includes subsampling in the x and y directions, and a large offset over the 50 pixel-wide WFC chip gap. Several astrodrizzle runs using different parameter values are used to determine the best settings for optimizing image sampling.
Summary of Steps
Run astrodrizzle with final_scale of 0.03 arcseconds using three final_pixfrac setting: 0.1. 0.6, and 1.0.
Evaluate the results to determine which parameters provide improved image resolution while preserving signal-to-noise to an acceptable level.
Data for this example are images of the galaxy NGC 4449 (Program 10585), obtained by the ACS/WFC using the F555W filter. These four images were obtained using primary and secondary two-point ACS-WFC-DITHER-LINE patterns. The primary dither pattern had a large offset to cover the 50 pixel-wide WFC chip, while the secondary dither pattern was a two-point subpixel dither at each point in the primary pattern. Offsets in the detector x and y direction are shown in Table 7.6 as POS TARG values in arcseconds.
Table 7.6: Summary of Images in this Example
Line Number

POS TARG are commanded offsets. The actual measured offsets would be slightly different.

These images are flat-field calibrated by calacs, and the flc.fits extension indicates they were corrected for CTE.
Figure 7.25: Dither Pattern Used for these Observations
Pattern includes subpixel sampling and large offset in y for covering the CCD gap.
Figure 7.26: Science (Top) and Data Quality Arrays (Bottom) of the First Image in the Association
The bottom image has been scaled to only show cosmic ray bits.
As described in Section 2.2, subpixel dithering has the potential to improve the resolution of the final combined image. Input pixels may be “shrunk” before being transformed to the subsampled output image. This is done to create a smaller input “footprint” in the output frame. The shrunken pixels, also called “drops,” rain down on the subsampled output frame where each shrunken input pixel is averaged into an output pixel with a weight proportional to the area of overlap between the “drop” and the output subsampled grid.
If the drop size is too small, not all output pixels will have data added to them from each input image. Care should be taken in selecting a drop size that is small enough to avoid degrading the image, but large enough to provide reasonably uniform coverage in the final output image. The drop size is controlled by the astrodrizzle parameter called final_pixfrac, which is the ratio of the linear size of the drop to the input pixel (before geometric distortion corrections). The level of subsampling for the output image is specified by the final_scale parameter.
Values for final_pixfrac and final_scale depend on the number of input images and the size of the shifts. In this example, the dither pattern provides some subsampling so the resolution can be improved. Ideally, the best scale is half that of the native scale, but to avoid degrading the PSF, users should not excessively decrease the value of the final_scale parameter.
For this example, a final_scale value of 0.03 arseconds/pixel is adopted, which is 60% the linear size of the native plate scale (0.05 arcseconds/pixel.). A series of experiments will be run to determine the best final_pixfrac value for the selected output image scale.
As suggested in the HST Dither Handbook, statistics performed on the drizzled weight image should yield a RMS/median value less than 0.2. This threshold controls the trade-off between improving image resolution versus increasing background noise due to pixel resampling.
A series of astrodrizzle commands, executed in PyRAF, can be written to create a test grid of different final_pixfrac values. Default values have been used for the parameters, except for final_pixfrac and final_scale. In the example below, three final_pixfrac values are used to illustrate its effect on the image quality: 1.0, 0.6, and 0.1.
Each astrodrizzle run produces three images; the science file (suffix drc_sci.fits), the weight map (drc_wht.fits), and the context image (drc_ctx.fits).
Figure 7.27 shows the central region of the science and weight images produced by final_pixfrac values of 1.0, 0.6, and 0.1. The science image with the smallest final_pixfrac, 0.1, shows a noisy background with hot pixels improperly removed. There are “holes” in the images where no input pixels fall into the output grid because they have been shrunk too small. Inspection of the weight map corresponding to the smallest final_pixfrac value shows many places with weights of zero, indicating that a final_pixfrac value of 0.1 is clearly too small, and was only included in this example for illustrative purposes.
Figure 7.27: Results for Three Tests with final_pixfrac Values of 1.0, 0.6, and 0.1
Left column corresponds to final_pixfrac =1.0, the middle column corresponds to a final_pixfrac = 0.6, and the right column was created using final_pixfrac = 0.1. The top row shows the weight maps, and the bottom row shows the science frames.
Weight statistics, (RMS/median) for each trial are obtained for a 500x500 pixel box in the center and lower right corner of each weight image. Statistics for the weight image created with final_pixfrac of 0.6 meet the general requirement of RMS/median < 0.2. A visual inspection of the weight image shows how the RMS varies over the field of view, due to changes in geometric distortion over the chip where input pixels cover different areas on the sky. (For more information, refer to the discussion on Pixel Area Maps at the ACS website.)
Determining the best solution for drizzle-combining images will also depend on the position of the target on the detector. For a compact source in the lower right corner of the image, one could set final_pixfrac to 0.5. For this example, however, objects are distributed over the entire field so a final_pixfrac of 0.6 is found to provide the best result, where the RMS/median does not exceed 0.2 on any portion of the detector.
Figure 7.28: Weight Map Statistics in a 500x500 Pixel Box at the Center and Lower Right Corner of the Image

Table of Contents Previous Next Index Print