Citation

Abstract

We illustrate Cramer–Rao lower bounds (CRLBs) on the root-mean-square (RMS) error in estimating the position of an image on a detector array. Several models of the image intensity distribution are treated. The case of Poisson arrivals and a Gaussian image profile, which was treated in [1], is extended to treat an unknown image mean and a skew-normal profile with unknown skew. We show the bounds are insensitive to knowledge of the mean while unknown skew can result in a loss of ≈1.5 dB. We show that a time-varying image profile, with pixel variance quadratic in the image mean, leads to a floor in the RMS error as a function of the mean.

Details

Volume
42-167
Published
November 15, 2006
Pages
1–17
File Size
236.4 KB