IRAC: Software: automuxstripe.pro

Before correction.	After correction of column-pulldown, muxbleed, and muxstriping.

This IDL routine corrects a "striping" feature associated with bright sources in channels 1 and 2. The stripe is usually confined to a single readout channel (IRAC has 4 readouts, thus, every fourth pixel in a row belongs to the same readout channel), occurs in a band across the image, and always results in a depressed bias level in the affected readout channel. This code automatically identifies the presence of a muxstripe and corrects it. The routine does the following:

1. Determines which channel has the stripe (if any) by finding the readout channel with a large number of anomalously low valued pixels relative to the rest of the image based on the noise in the image.
2. Scans down that channel, searching for three consecutive pixels N-sigma below the mean, signifying the start of the stripe.
3. Continues on, searching for consecutive pixels in the same readout channel that are within N-sigma of the mean, thus signifying the end of the stripe.
4. Once the stripe is identified, break the image into pieces and force the DC level region of the readout channel within the stripe to match the rest of the image.

To run the code within IDL, type a command like

automuxstripe,'in.fits','out.fits',thresh=3,/VERBOSE,/CHECKIMAGE

See the .pro file for explanations and further qualifications (for example, muxbleed and column pulldown have to be corrected before using this code).

This routine originated to fix muxstriping in images that consisted of a single bright object on a flat background. It has also been successfully adapted for the SWIRE and S-COSMOS projects. In the latter cases it was modified to make it process entire lists at a time, and to use prior knowledge of the field contents based on 2MASS. Users should be aware that in many circumstances this code will fail, and in particular complex backgrounds may confuse it. Users should consider this code to be demonstrative of at least one way of fixing the muxstripe problem. Other strategies clearly exist:

• Rather than using automatic detection, a user could supply start and stop points based on visual inspection.
• A truth image (either the finished mosaic, or some other image) could be used to model the background to allow for a more accurate derivation of statistics.
• Prior knowledge (such as a 2MASS source list) could be used to predict which stars will trigger muxstriping. However, the end of the stripe is usually difficult to predict, and this may not work well.
• A 2MASS source list could be used to determine how often to apply the algorithm (a technique used by S-COSMOS and SWIRE).

This software is not considered part of the pipeline and no installation support from the SSC should be expected. We are however very interested in your feedback in order to improve or replace the algorithm and are glad to work with you in test cases and analysis. Please send your comments to help@spitzer.caltech.edu.

The IDL procedure automuxstripe.pro.