Abstract
Recently, numerous tensor singular value decomposition (t-SVD)-based tensorrecovery methods have shown promise in processing visual data, such as colorimages and videos. However, these methods often suffer from severe performancedegradation when confronted with tensor data exhibiting non-smooth changes. Ithas been commonly observed in real-world scenarios but ignored by thetraditional t-SVD-based methods. In this work, we introduce a novel tensorrecovery model with a learnable tensor nuclear norm to address such achallenge. We develop a new optimization algorithm named the AlternatingProximal Multiplier Method (APMM) to iteratively solve the proposed tensorcompletion model. Theoretical analysis demonstrates the convergence of theproposed APMM to the Karush-Kuhn-Tucker (KKT) point of the optimizationproblem. In addition, we propose a multi-objective tensor recovery frameworkbased on APMM to efficiently explore the correlations of tensor data across itsvarious dimensions, providing a new perspective on extending the t-SVD-basedmethod to higher-order tensor cases. Numerical experiments demonstrated theeffectiveness of the proposed method in tensor completion.