SNR-dependent stopping criterion for improved support recovery

Develop and validate a stopping criterion for multiscale generalized sparse plane wave decomposition that depends on the signal-to-noise ratio and improves support-recovery performance.

Background

The multiscale GSPWD algorithm terminates either after a prescribed number of iterations when the sparsity level is known or when a relative residual-energy threshold is reached. In the simulations, a fixed threshold is used, and the results show that performance depends substantially on noise level. The paper therefore identifies the choice of stopping rule as an unresolved methodological issue: an SNR-adaptive criterion might better distinguish meaningful plane-wave components from noise and residual modeling error, potentially reducing over-selection or under-selection.

The problem concerns the design and evaluation of an SNR-dependent termination rule specifically for improving support recovery in the proposed sparse decomposition procedure.