TL
TL;DR
The paper presents a unified, flexible framework for the tasks of audio inpainting, declipping, and dequantization that is extended to cover analogous degradation models in a transformed domain, e.g. quantization of the signal's time-frequency coefficients.