Advanced Image Processing

Many of my research efforts rely on developing new techniques in processing of hyperspectral and time-series imagery to reveal features buried by contamination from noise, background sources, or foreground sources.

An example of n=5 kmao-optimized apertures from the Kepler Titan dataset.

In 2019, I developed a technique called k-Means Aperture Optimization (kmao) for extracting high-precision photometry from data containing blended, moving saturated sources with time-varying PSFs and scattered light. I designed it after acquiring a Kepler image time-series of Titan moving through Saturn’s scattered light. Instead of identifying a single optimum photometric aperture, kmao optimizes a small set of apertures. These apertures each apply to a unique sub-set of the target images that have similar properties. The assigment of images to these sets is done via k-means clustering on the target pixel files. The kmao-processed photon-tagged time series is illustrated below, with cyan indicating star-sourced light, magenta indicating Saturn-sourced light, and yellow indicating Titan-sourced light. White pixels show where the detector saturated.

Even when up to 60% of the photons recorded in the aperture were from the time-varying scattered light of Saturn, we were able to extract photometry from Titan with better than 0.2% precision.

More details about kmao and the Kepler Titan dataset can be found in this blog post or in our 2019 PASP paper, “k-Means Aperture Optimization Applied to Kepler K2 Time Series Photometry of Titan.”