Spatio-Temporal Signal Recovery via Low-Rankness and Smoothness of Difference Features

Spatio-temporal signal recovery has attracted considerable research interest, with emphasis on exploiting the low-rankness and smoothness characteristics of signals. This paper introduces a method termed STELLAR (spatio-temporal signal recovery via low-rankness and smoothness of difference features), which fuses global low-rankness and local spatio-temporal smoothness of signals in a unified manner. Specifically, we apply the nuclear norm to difference features after three spatiotemporal difference transformations of the signal tensor, thereby simultaneously quantifying the low-rankness and smoothness. This yields a convex formulation of STELLAR, which is efficiently solved using the alternating direction method of multipliers (ADMM). Extensive experiments demonstrate that STELLAR significantly outperforms many state-of-the-art methods while maintaining superior computational efficiency. The STELLAR code is available at https://github.com/IAmEzreal/STELLAR.git.