Abstract:We consider the problem of recovering two unknown vectors, $\boldsymbol{w}$ and $\boldsymbol{x}$, of length $L$ from their circular convolution. We make the structural assumption that the two vectors are members of known subspaces, one with dimension $N$ and the other with dimension $K$. Although the observed convolution is nonlinear in both $\boldsymbol{w}$ and $\boldsymbol{x}$, it is linear in the rank-1 matrix formed by their outer product $\boldsymbol{w}\boldsymbol{x}^*$. This observation allows us to recast the deconvolution problem as low-rank matrix recovery problem from linear measurements, whose natural convex relaxation is a nuclear norm minimization program. We prove the effectiveness of this relaxation by showing that for "generic" signals, the program can deconvolve $\boldsymbol{w}$ and $\boldsymbol{x}$ exactly when the maximum of $N$ and $K$ is almost on the order of $L$. That is, we show that if $\boldsymbol{x}$ is drawn from a random subspace of dimension $N$, and $\boldsymbol{w}$ is a vector in a subspace of dimension $K$ whose basis vectors are "spread out" in the frequency domain, then nuclear norm minimization recovers $\boldsymbol{w}\boldsymbol{x}^*$ without error. We discuss this result in the context of blind channel estimation in communications. If we have a message of length $N$ which we code using a random $L\times N$ coding matrix, and the encoded message travels through an unknown linear time-invariant channel of maximum length $K$, then the receiver can recover both the channel response and the message when $L\gtrsim N+K$, to within constant and log factors.

Generalized variational framework with minimax optimization for parametric blind deconvolution

Multi-frame Blind Deconvolution Using Sparse Priors

Blind Deconvolution Using an Improved L0 Sparse Representation

Identifiability of Multichannel Blind Deconvolution and Nonconvex Regularization Algorithm

Lp (1≤p≤∞) BLIND ADAPTIVE DECONVOLUTION AND ITS APPLICATIONS

Non-blind image deconvolution using natural image gradient prior

L∞ Blind Deconvolution for the Generalized AR Model.

A Deep Variational Bayesian Framework for Blind Image Deblurring

Rapid, Robust, and Reliable Blind Deconvolution Via Nonconvex Optimization

A Novel Approach Based on Differential Evolution for Blind Deconvolution.

Blind Image Deconvolution Using Variational Deep Image Prior

Fundamental Limits of Blind Deconvolution Part I: Ambiguity Kernel

Blind Deconvolution using Convex Programming

Discriminative learning of iteration-wise priors for blind deconvolution

Blind Image Deblurring Using Spectral Properties of Convolution Operators

Learning Deep Optimizer For Blind Image Deconvolution

Semiparametric Approach to Multichannel Blind Deconvolution of Nonminimum Phase Systems

Blind Image Deblurring with Local Maximum Gradient Prior.

Multi-Scale Blind Motion Deblurring Using Local Minimum

Applications of multiscale hierarchical decomposition to blind deconvolution

Blind Image Deconvolution Assisted By Robust Non-Gaussian Priors