Ronald Aung, Air Force Institute of Technology; Stephen Cain, Air Force Institute of Technology
Keywords: Imaging through Turbulence, Image Restoration, Dim Object Detection, Closely Spaced Objects
Abstract:
The problem of directly imaging stellar objects in close proximity to bright ones remain a topic of high interest and a challenging one. Although non-imaging techniques have been successful in identifying the presence of planets, the desire to see them directly still remains in order to verify models used in these searches. Transit methods have been successful in the direct imaging of planets but afforded limited opportunities for detecting planets due to the relative geometry among the space surveillance telescope and the planets. In this research, we introduce a method that allows for direct imaging of planetary bodies during non-transit times, which would potentially allow for more planets to be discovered in a short period of time. The proposed algorithm does not negate the need for a coronagraph or other techniques for improving the contrast between a dim star and a bright one. Since the proposed algorithm works on principles completely independent of the functioning of a coronagraph, it can be used in conjunction with one to improve the detection of a dim star around a bright one above what could be obtained from these methods used separately.
This research describes a two-step blind deconvolution algorithm, which is built on the method proposed by Schulz [1] in the first step and that by Cain [2] in the second step. The first step is primarily used in estimating the bright star, while improving the estimates of the point spread function (PSF). In the second step, we introduce a new algorithm that separates the astronomical image function into three sets, which are the primary bright star, the unknown neighborhood system function around the bright star, and the background light and dark current measured during the acquisition process. With these three sets, the new algorithm uses an iterative Expectation Maximization (EM) algorithm similar to the one employed by Schulz [1] and Shepp and Vardi [3]. The EM algorithm was brought to widespread attention of statistical community by Dempster, Laird, and Ruben, and subsequently corrected some errors in that work regarding the convergence by Wu [4][6]. Based on our analysis, Schulzs algorithm works very well for resolving closely spaced stellar objects of similar intensity, such as binary stars. However, when the intensity of the secondary star is reduced, the ability of the algorithm to detect and resolve the dim one also decreases.
With the knowledge of the bright star and the PSF estimates from the first step, this research explores the potential performance improvements of additional signal processing with the new algorithm in the second step. This research compares the performance gained using two metrics, which are the ability to detect the dim star and the resolution between the dim star and the bright one. For this research, we make the assumption that we have the knowledge of the optical and detector models. In addition, since the focal plane array can measure the intensity of astronomical images without the phase information directly, this research uses a phase retrieving algorithm to recover the phase aberration with post-processing [7], [8]. This research applies Gerchberg-Saxton phase retrieval algorithm in both steps [9]. The scope of this research includes analysis through modeling and simulation using computer simulated data. MATLAB® built-in functions are used throughout the analysis. Comparisons between the new algorithm in the second step and the Schulzs method in the first step are presented, showing the degree to which the dim star in close proximity to the bright one can be detected and resolved.
REFERENCES
[1] T. J. Schulz, Multiframe Blind Deconvolution of Astronomical Images, J. Opt. Soc. Am., vol. 10, no. 5, p. 1064, 1993.
[2] S. C. Cain, Super-Resolution Imaging via Expectation-Maximization Estimation of Near Stellar Neighborhoods, in Proc. SPIE, 2018, vol. 10772.
[3] L. A. Shepp and Y. Vardi, Maximum Likelihood Reconstruction Emission Tomography, IEEE Trans. Med. Imaging, vol. MI1, no. 2, pp. 113122, 1982.
[4] R. B. Millar, Maximum Likelihood Estimation and Inference: with examples in R, SAS and ADMB. Chichester, West Sussex: Wiley, 2011.
[5] A. P. Dempster, N. M. Laird, and D. B. Rubin, Maximum Likelihood from Incomplete Data via the EM Algorithm, vol. 39, no. 1. 1977.
[6] C. F. Wu, On the Convergence Properties of the EM Algorithm, Ann. Stat., vol. 11, no. 1, pp. 95103, 2013.
[7] J. R. Fienup, Phase Retrieval Algorithms: a comparison, Appl. Opt., vol. 21, no. 15, pp. 27582769, 1982.
[8] J. C. Zingarelli and S. C. Cain, Phase Retrieval and Zernike Decomposition using Measured Intensity Data and the Estimated Electric Field, Appl. Opt., vol. 52, no. 31, pp. 74357444, 2013.
[9] R. W. Gerchberg and W. O. Saxton, A Practical Algorithm for the Determination of the Phase from Image and Diffraction Plane Pictures, Optik (Stuttg)., vol. 35, pp. 237246, 1972.
Date of Conference: September 17-20, 2019
Track: Adaptive Optics & Imaging