DETECTION AND CHARACTERIZATION OF EXOPLANETS AND DISKS USING PROJECTIONS ON KARHUNEN-LOÈVE EIGENIMAGES

Abstract

We describe a new method to achieve point spread function (PSF) subtractions\nfor high- contrast imaging using Principal Component Analysis (PCA) that is\napplicable to both point sources or extended objects (disks). Assuming a\nlibrary of reference PSFs, a Karhunen-Lo`eve transform of theses references is\nused to create an orthogonal basis of eigenimages, on which the science target\nis projected. For detection this approach provides comparable suppression to\nthe Locally Optimized Combination of Images (LOCI) algorithm, albeit with\nincreased robustness to the algorithm parameters and speed enhancement. For\ncharacterization of detected sources the method enables forward modeling of\nastrophysical sources. This alleviates the biases in the astrometry and\nphotometry of discovered faint sources, which are usually associated with LOCI-\nbased PSF subtractions schemes. We illustrate the algorithm performance using\narchival Hubble Space Telescope (HST) images, but the approach may also be\nconsidered for ground-based data acquired with Angular Differential Imaging\n(ADI) or integral-field spectrographs (IFS).\n