Abstract
It has been shown that expectation-maximization (EM) can be applied to a maximum-a posteriori (MAP) formulation of the image restoration problem, resulting in a nonlinear iterative restoration algorithm. This MAP/EM algorithm has been shown to be effective in the restoration and super resolution of point objects. When applied to ex tended objects such as planets, however, the algorithm produces ringing artifacts near edges in the object. We show that such artifacts can be overcome by decomposing the image into two terms, a background and a foreground. The background term is used to remove large scale variations in the data such that the foreground term retains primarily edge information. The MAP/EM algorithm then operates on the foreground term to produce the deblurring. This decomposition effectively generalizes the positivity constraint beyond a zero-level surface. We describe our means of performing the decomposition and show both simulated and actual results.
© 1991 Optical Society of America
PDF ArticleMore Like This
Peter Nisenson
MQQ2 OSA Annual Meeting (FIO) 1991
Donald. L. Snyder, Timothy J. Schulz, and Abed M. Hammoud
WC1 Signal Recovery and Synthesis (SRS) 1992
Timothy J. Schulz
MBBB.2 OSA Annual Meeting (FIO) 1993