A likelihood based method for compressive signal recovery under Gaussian and saturation noise

Shuvayan Banerjee, Sudhansh Peddabomma, Radhendushka Srivastava, Ajit Rajwade

Abstract: Highlights•We present a novel likelihood based estimator for compressive reconstruction from measurements that are saturated and are also corrupted with Gaussian noise.•We prove important properties of this estimator and derive performance bounds for the associated reconstruction error, which we show are provably superior to those associated with an estimator which would just discard saturated measurements.•We present numerical results for reconstruction of synthetic signals and images from compressive measurements using our estimator, and demonstrate superior numerical performance in comparison to several existing estimators. We also demonstrate an application of our estimator in audio de-clipping.

External IDs:dblp:journals/sigpro/BanerjeePSR24