A SURE Approach for Digital Signal/Image Deconvolution Problems

In this paper, we are interested in the classical problem of restoring data degraded by a convolution and the addition of a white Gaussian noise. The originality of the proposed approach is two-fold. Firstly, we formulate the restoration problem as a nonlinear estimation problem leading to the minimization of a criterion derived from Stein’s unbiased quadratic risk estimate. Secondly, the deconvolution procedure is performed using any analysis and synthesis frames that can be overcomplete or not. New theoretical results concerning the calculation of the variance of the Stein’s risk estimate are also provided in this work. Simulations carried out on natural images show the good performance of our method w.r.t. conventional wavelet-based restoration methods.

💡 Research Summary

The paper tackles the classic inverse problem of restoring signals or images that have been blurred by a convolution and corrupted by additive white Gaussian noise. Its novelty lies in two main aspects. First, the authors recast the deconvolution task as a nonlinear estimation problem and adopt Stein’s unbiased risk estimate (SURE) as the objective function. Because SURE provides an unbiased estimate of the mean‑squared error (MSE) without requiring knowledge of the true signal, it serves as a principled, data‑driven criterion for selecting regularization parameters or threshold levels. Second, the proposed framework is agnostic to the choice of analysis and synthesis frames; any pair of frames—orthogonal, redundant, or over‑complete—can be employed. This flexibility allows the use of traditional wavelets, dual‑tree wavelets, over‑complete DCT dictionaries, or even learned dictionaries.

Mathematically, let y = h * x + n denote the observed image, where h is the blur kernel, x the unknown clean image, and n ∼ N(0,σ²I) the Gaussian noise. An analysis operator Φ maps x to a coefficient vector α = Φx, while a synthesis operator Ψ reconstructs an estimate (\hat{x}=Ψβ) from coefficient β. The estimator β = f(α;θ) is defined through a smooth, element‑wise non‑linear function (e.g., soft‑thresholding) parameterized by θ (thresholds, scaling factors, etc.). The SURE risk for a given θ can be expressed in closed form as
\