“…Generally, HSI-MSI fusion approaches can be categorized as three classes, i.e., low-rank based- [7], [8], [9], [10], [11], sparse based- [12], [13], [14], and deep-learning basedmethods [15], [16], [17]. Instead of modeling HSI directly at the pixel level, low-rank based methods usually adopt matrix or tensor decomposition, supplemented by appropriate constraints, to search the potential HR-HSI.…”