Imtrandetect: a new tool/methodology for detecting astronomical transients
from large image-data streams down to low S/N

Frank Masci, Douglas Hoffman, and Ashish Mahabal

One of the challenges in current and future time-domain surveys is to identify
reliable candidate transient and variable sources for further follow-up and
classification (if possible). We describe a new methodology for optimally
detecting transients from multi-epoch image data, with an emphasis on minimizing
contamination from instrumental artifacts and glitches. The method is based on
collapsing prior-masked image stacks into various metric-statistic images for
later thresholding and object detection. The stacking is performed in moving
windows along the time-ordered image sequence where window lengths are tuned
to maximize the sensitivity of the metrics for detecting intermittent local
transient behavior above the baseline noise traced by the full time-series
of each pixel signal. We show example light curves generated by the tool,
from testing on image data containing known transients and variables from
the Catalina Real-time Transient Survey (CRTS) and the Wide-field Infrared
Survey Explorer (WISE). We also show the light curves of some new interesting
candidate transients and variables found in these test fields.