Current plans call for the first Terrestrial Planet Finder mission, TPF-C, to be a monolithic space telescope with a coronagraph for achieving high contrast. The coronagraph removes the diffracted starlight allowing the nearby planet to be detected. In this paper, we present a model of the planet measurement and noise statistics. We utilize this model to develop two planet detection algorithms, one based on matched filtering of the PSF and one using Bayesian techniques. These models are used to formulate integration time estimates for a planet detection with desired small probabilities of false alarms and missed detections.

The past five years has seen a surge in research and innovative ideas for the imaging of extrasolar planets, particular terrestrial ones. We expect that within the next decade a space observatory will be launched with the objective of imaging earthlike planets. Because of the limited lifetime of such a mission and the large number of potential targets, integration time is a critical parameter. In fact, integration time is the primary metric in evaluating various design approaches for the high contrast imaging system. In this paper we present a new approach to determining the existence of a planet in an observed system using Bayesian hypothesis testing. Rather than perform photometry, or rely on vision to determine the existence of a planet, this approach evaluates the image plane data statistically under certain assumptions about the prior probability distributions. We show that extremely high confidence can be achieved in substantially shorter integration times than conventional photometric methods.