Partially Linear Estimation With Application to Sparse Signal Recovery From Measurement Pairs

Abstract

We address the problem of estimating a random vector from two sets of measurements and , such that the estimator is linear in . We show that the partially linear minimum mean-square error (PLMMSE) estimator does not require knowing the joint distribution of and in full, but rather only its second-order moments. This renders it of potential interest in various applications. We further show that the PLMMSE method is minimax-optimal among all estimators that solely depend on the second-order statistics of and . We demonstrate our approach in the context of recovering a signal, which is sparse in a unitary dictionary, from noisy observations of it and of a filtered version. We show that in this setting PLMMSE estimation has a clear computational advantage, while its performance is comparable to state-of-the-art algorithms. We apply our approach both in static and in dynamic estimation applications. In the former category, we treat the problem of image enhancement from blurred/noisy image pairs. We show that PLMMSE estimation performs only slightly worse than state-of-the art algorithms, while running an order of magnitude faster. In the dynamic setting, we provide a recursive implementation of the estimator and demonstrate its utility in tracking maneuvering targets from position and acceleration measurements.Index Terms-Bayesian estimation, deblurring, denoising, minimum mean-square error, target tracking.

LMMSE Filtering in Feedback Systems With White Random Modes: Application to Tracking in Clutter

“…Among these we mention [2] and [3] that considered estimation with uncertain observations, [7] that derived a Kalman filter-like (KF) algorithm for a JLS with independently switching modes and uncorrelated matrices within each time step, and [8] that derived an LMMSE scheme for a Markov JLS by means of state augmentation. In addition, in some cases, parts of the state may be estimated optimally while others in a linear optimal manner, as was shown in [9].…”

Recent Progress in Image Deblurring

“…To be clear, the noisy image can first be denoised to produce a nearly sharp image which can then be used as a guideline or a constraint in the deblurring process. A recent approach by Michaeli et al (2012) has exploited this strategy by proposing the partially linear MMSE (PLMMSE) estimator. Denoting the denoised image C as a constraint on the estimation of W and v in (3.29), PLMMSE is…”

Semi-supervised single- and multi-domain regression with multi-domain training

“…where a a a : R M 1 → R N is an arbitrary function and B B B ∈ R N×M 2 is some matrix. It was shown in [16] that the solution to this particular case is given by…”