Multichannel blind deconvolution of polarimetric imagery

Abstract

A maximum likelihood blind deconvolution algorithm is derived for incoherent polarimetric imagery using expectation maximization. In this approach, the unpolarized and fully polarized components of the scene are estimated along with the corresponding angles of polarization and channel point spread functions. The scene state of linear polarization is determined unambiguously using this parameterization. Results are demonstrated using laboratory data.

Multiframe-multichannel blind deconvolution for polarimetric imagery

“…The results in [1] demonstrate the viability of the polarimetric blind deconvolution algorithm. What remains then, is to determine how the algorithm compares to the alternatives.…”

“…This paper further develops the multichannel polarimetric blind deconvolution algorithm originally presented by the author in [1]. In section 4, this algorithm is compared to blind deconvolution of the individual imaging polarimeter channels.…”

“…The polarimetric blind deconvolution algorithm derived in [1] specifically addresses the case where all quantities of interest are estimated from a single frame for each polarization sensitive channel. In this derivation, these data describe a log-likelihood function that is maximized across the parameter set of interest.…”

“…The difference between these two approaches lies in how the PSFs are estimated. Since there is no obvious advantage to choosing one PSF estimator over the other, the Schulz PSF estimator, which is also the PSF estimator in [1], is used here for consistency.…”

“…In [1], a maximum likelihood (ML) multichannel blind deconvolution algorithm using expectation maximization is derived for incoherent polarimetric imagery. The purpose of this algorithm is to estimate the true linear polarization state of a target scene in the presence of photon noise and an unknown atmospheric path.…”