The MultiDrizzle Handbook
7.1 Interpreting your results
MultiDrizzle ties together a fairly substantial set of algorithms, each designed to accomplish a different task, and as such has a large parameter set. While the functions have been set to default parameters which should produce good results for a wide range of data, there are times when they must be fine-tuned to achieve the best results. Along the way it is possible to run through the entire process without software error but end up with a product which is not of the optimum science quality. The following section attempts to outline possible outcomes and their probable solutions.
7.1.1 General Issues
- When I display the weight map generated for my final drizzled image (or any of the single drizzled images as well), there seem to be representations of the objects in my science image. What's going on?
There are several issues which you should examine:
- Most likely you need to refine the parameters that are set during the driz_cr step which detects cosmic rays in each of your science frames. This step uses the science image, a model of that image (the blotted median image), and the absolute derivative of the model image to locate cosmic rays. It uses the scaling, readnoise and gain specified by the user as well as the background level specified in the header of the image as part of its noise comparison. Often times the cores of saturated or very bright objects will be flagged as cosmic rays if the scaling is improper.
- You should double check the alignment of your images. You can do this by visually blinking between your science image and your blotted image. If the images are not well aligned when the median is created then the objects in the blotted image will not fall in the same place as the original science image - leading the incorrectly flagged pixels.
- I've calculated refined offsets between my science images but when I supply MultiDrizzle with the delta shifts the final image is not aligned properly.
- Make sure that you have calculated your shifts and rotations using the same frame of reference that the MultiDrizzle software is expecting. The most important thing is that all your shifts are stated relative to the reference image that you chose. You can specify the reference image that was used to calculate the shifts inside the delta shift file.
- When I look at my output drizzled image, there seem to be obvious seams between each pointing.
- This could be a sky subtraction problem. Sky in this case can be somewhat of a misnomer, what it means is that the background levels in all your input images have not been normalized to each other. It is important to have a good background subtraction before drizzling your images so that proper cosmic ray rejection can be performed and so that you are not adding unneccesary noise to your combined output image. You might also be drizzling images with vastly different S/N ratios which could have the same affect.
- This could also be a problem with the shifts between your images. If your shifts are close, but not precise, you may see ringing around your objects, edges between observations and distorted or smeared star profiles. These will vary depending on the morphology of your field.
7.1.2 Instrument Specific Information
These are questions you might encounter as you drizzle your data and are related to specific instrumental characteristics.
- I just retrieved ACS data from the archive, but the images are very noisy and there are stripes all over the fields, which is not common in the ACS data.
The correlated noise "stripes" you see are most prominent when low S/N narrowband exposures are drizzled. When the significant ACS distortion correction is applied to such data, it can impose this Moire-like beat pattern in the noise. The best way to mitigate against this is to employ a dither pattern or otherwise combine more frames from overlapping observations, so that the correlated noise will be out of phase and cancel each other in the combined output. You can run MultiDrizzle yourself to combine any number of frames. Note that in combining data from different visits (which employed different guide stars), you will likely need to manually register the frames (measure and apply shifts).
If the data is undithered, and more frames are not available for combination, you might want to try using the lanczos3 kernel for the final drizzle step in MultiDrizzle. It can help suppress the correlated noise, but note that it doesn't perform well in the presence of artifacts such as hot pixels and cosmic rays. So if you are not combining enough frames to reject all such artifacts, you will notice some "ringing" (a halo of negative pixels) around the sharp edges of any residual artifacts.
Space Telescope Science Institute
Voice: (410) 338-1082