The MOST (Microvariability & Oscillations of STars) satellite obtains ultraprecise photometry from space with high sampling rates and duty cycles. Astronomical photometry or imaging missions in low Earth orbits, like MOST, are especially sensitive to scattered light from Earthshine, and all these missions have a common need to extract target information from voluminous data cubes. They consist of upwards of hundreds of thousands of two-dimensional CCD frames (or sub-rasters) containing from hundreds to millions of pixels each, where the target information, superposed on background and instrumental effects, is contained only in a subset of pixels (Fabry Images, defocussed images, mini-spectra). We describe a novel reduction technique for such data cubes: resolving linear correlations of target and background pixel intensities. This stepwise multiple linear regression removes only those target variations which are also detected in the background. The advantage of regression analysis versus background subtraction is the appropriate scaling, taking into account that the amount of contamination may differ from pixel to pixel. The multivariate solution for all pairs of target/background pixels is minimally invasive of the raw photometry while being very effective in reducing contamination due to, e.g., stray light. The technique is tested and demonstrated with both simulated oscillation signals and real MOST photometry.
Download