Quantitative Performance of the Mopex Multi-Frame Outlier-Detection Algorithm

Abstract

The Mopex software is used at the Spitzer Science Center (SSC) to produce co-added and mosaicked images from sets of individually processed Spitzer images. Until now, quantitative studies of the performance of Mopex's outlier-detection methods had never been performed. This particular study focuses only on Mopex's multiframe outlier-detection algorithm, and future studies are still needed to characterize its so-called box and dual methods. The performance of the multiframe method varies with two adjustable parameters, ΓTOP and ΓRM. For a given ΓTOP value, we computed the completeness (C) and reliability (R) of the outlier detection for 101 discrete values of ΓRM uniformly distributed in the full range of possible settings for this parameter, which lie continuously between 0 and 1, inclusive. We characterized the C and R performance as a function of ΓRM in this manner for ΓTOP values of 2, 2.5, 3, 5, and 10 for image data in all four IRAC channels (infrared passbands). Not surprisingly, the performance for IRAC channel 3 is relatively poor because the image data for this channel are markedly noisier. The best performance was obtained for a ΓTOP value of 3, and this applies to all four channels. Generally, setting ΓRM low will maximize completeness at the expense of reliability, and vice versa for setting ΓRM high. For example, for IRAC channel 1 and ΓTOP = 3, setting ΓRM = 0.3 gives C = 83% and R = 41%, and setting ΓRM = 0.8 gives C = 52% and R = 86%.

Files

Additional details