[Paper Review] On the Stable Resolution Limit of Total Variation Regularization for Spike Deconvolution

Maxime Ferreira Da Costa, Yuejie Chi|arXiv (Cornell University)|Oct 3, 2019

Sparse and Compressive Sensing TechniquesEngineering44 references21 citations

TL;DR

This paper establishes a precise, PSF-dependent separation criterion for stable support recovery of two closely spaced point sources using the Beurling-LASSO estimator with total variation regularization. It proves that when the sources are separated by more than a threshold derived solely from the point spread function (PSF), the estimator guarantees exact support recovery without spurious or missing spikes, even under noise. The result extends to multi-source settings with well-separated and closely located sources.

ABSTRACT

The stability of spike deconvolution, which aims at recovering point sources from their convolution with a point spread function (PSF), is known to be related to the separation between those sources. When the observations are noisy, it is critical to ensure support stability, where the deconvolution does not lead to spurious, or oppositely, missing estimates of the point sources. In this paper, we study the resolution limit of stably recovering the support of two closely located point sources using the Beurling-LASSO estimator, which is a convex optimization approach based on total variation regularization. We establish a sufficient separation criterion between the sources, depending only on the PSF, above which the Beurling-LASSO estimator is guaranteed to return a stable estimate of the point sources, with the same number of estimated elements as that of the ground truth. Our result highlights the impact of PSF on the resolution limit in the noisy setting, which was not evident in previous studies of the noiseless setting. Towards the end, we show that the same resolution limit applies to resolving two close-located sources in conjunction of other well-separated sources.

Motivation & Objective

• To determine the minimal separation distance between two closely located point sources that ensures stable support recovery under noisy conditions.
• To quantify how the point spread function (PSF) influences the resolution limit in spike deconvolution with total variation regularization.
• To establish a sufficient separation criterion based only on the PSF that guarantees the Beurling-LASSO estimator recovers the correct number of sources.
• To extend the stability result to scenarios with multiple sources, including both well-separated and closely located pairs.
• To rigorously analyze the asymptotic behavior of the estimator’s dual certificate and verify non-degeneracy conditions for support stability.

Proposed method

• Uses the Beurling-LASSO estimator, a convex optimization framework based on total variation regularization of signed measures.
• Models the observation as a convolution of point sources with a band-limited PSF, followed by uniform sampling of its Fourier transform.
• Introduces a normalized measure on the torus to simplify the analysis and exploit shift-invariance of the measurement operator.
• Derives a dual certificate via the inverse of the Gram matrix of the sensing operator, analyzing its asymptotic behavior as the number of samples N increases.
• Applies the Riemann-Lebesgue lemma and asymptotic analysis of Fourier kernels to show that off-diagonal blocks of the Gram matrix converge to zero.
• Verifies the non-degenerate source condition by proving that the dual certificate decays away from source locations and has negative curvature at source positions in the limit N → ∞.

Experimental results

Research questions

• RQ1 What is the minimal separation between two point sources that guarantees stable support recovery using total variation regularization in the presence of noise?
• RQ2 How does the point spread function (PSF) affect the resolution limit in noisy spike deconvolution?
• RQ3 Can the Beurling-LASSO estimator stably recover the support of two closely spaced sources when other well-separated sources are present?
• RQ4 Under what conditions does the dual certificate of the Beurling-LASSO estimator satisfy the non-degeneracy condition required for support stability?
• RQ5 Is the resolution limit dependent only on the PSF, and not on the amplitudes or phases of the sources?

Key findings

• The paper establishes a sufficient separation criterion between two point sources that depends only on the PSF, ensuring stable support recovery via the Beurling-LASSO estimator under noisy conditions.
• The resolution limit is determined by the decay properties of the Fourier transform of the PSF, specifically through the behavior of the kernel function K(τ) and its derivatives.
• As the number of measurements N increases, the off-diagonal blocks of the Gram matrix of the sensing operator converge to zero, ensuring that the dual certificate concentrates at source locations.
• The dual certificate satisfies the non-degeneracy condition in the limit N → ∞, which guarantees that the Beurling-LASSO estimator recovers the correct number of sources.
• The support stability result holds even when the two closely spaced sources are embedded in a mixture with other well-separated sources.
• The analysis confirms that the resolution limit is fundamentally governed by the PSF’s spectral characteristics, not by the signal amplitudes or phases.