This award will support the development of innovative image analysis algorithms for the direct detection of faint exoplanets in the presence of scattered starlight from the host star. The research team will try a new interdisciplinary approach, using ideas from computer vision experiments, with the goal of reducing systematic errors resulting from incomplete modeling of background and scattered light.
The work is expected to result in the creation of a flexible pipeline for the analysis of exoplanets and other objects. Codes from this project to the public will be released under open-source licenses.
This grant supported the investigation of a new type of software to aid in the direct imaging and study of planets orbiting stars other than the Sun. The funds were used primarily to support a research scientist at AMNH and resulted in new cross-discipline work with a computer vision expert at NYU, seeding astronomy with new techniques. Though thousands of exoplanets are now known, only a handful have been directly imaged, with very little spectral information extracted. The current best hope for routine direct detection is with advanced coronographs combined with very high-order adaptive optics (AO) and integral field spectrographs (IFS). These are high dynamic-range imagers that block out light from a very bright primary star to make it possible to detect and measure far fainter companions. In real systems a fraction of the primary light is scattered and diffracted by both the atmosphere and internal optics. The result is incomplete starlight removal and an image covered with speckles. Although this unblocked light is a tiny fraction of the primary stellar light, it can still be far brighter than the target exoplanet at every location in the focal plane. In high-contrast astronomy speckles have to be controlled with exquisite optics or removed in data processing, or both. A number of techniques for mitigating speckle noise in coronagraphic images have been developed over the past few years, but none of them were able to reach the physical limit of the photon shot noise. Speckle noise is typically 100 times brighter than this physical limit. Forging a new interdisciplinary approach, using ideas from computer vision and computer science, we developed a new technique for removing speckles, based on real data from the currently operational Project 1640 spectroscopic coronagraph at Palomar Observatory. The software was developed and tested and allows for much fainter objects next to stars to be found and analyzed than ever before, reaching the physical limit to within a factor of 2 rather than the much worse factor of 10 to 30 that other techniques yield. The software is in regular use now and is available for use by any other research group in the world. We published one paper on this new algorithm and a second one is in preparation.
