Characterization of Unstable Pixels Using a Mixture Model: Application to HST WFC3 IR

TL;DR

This paper introduces a Bayesian method to identify unstable pixels in infrared datasets, improving the accuracy of pixel stability characterization for HST WFC3 IR and future space telescopes.

Contribution

A new Bayesian approach for detecting unstable pixels that outperforms previous methods in purity and completeness, applicable to multiple datasets including HST and future missions.

Findings

Better identification of unstable pixels with higher purity.

More complete unstable pixel lists compared to previous updates.

Applicable to datasets with dithering, relevant for upcoming space telescopes.

Abstract

Many IR datasets are taken with two dithers per filter, complicating the automated recognition of pixels with unstable response. Much data from the HST cameras NICMOS and WFC3 IR fall into this category, and future JWST and WFIRST data are likely to as well. It is thus important to have an updated list of unstable pixels built from many datasets. We demonstrate a simple Bayesian method that directly estimates the fraction of the time the output of each pixel is unstable. The last major update for WFC3 IR was a 2012 instrument science report (ISR WFC3 2012-10, Hilbert 2012), so we compute a new list. Rather than reproduce the old analysis on newer data, we use our new method. By visual inspection, our method identifies unstable pixels with better purity and completeness.