Hyperspectral Image Recovery Constrained by Multi-Granularity Non-Local Self-Similarity Priors

Hyperspectral image (HSI) recovery, as an upstream image processing task,holds significant importance for downstream tasks such as classification,segmentation, and detection. In recent years, HSI recovery methods based onnon-local prior representations have demonstrated outstanding performance. However,these methods employ a fixed-format factor to represent the non-local self-similaritytensor groups, making them unable to adapt to diverse missing scenarios. To addressthis issue, we introduce the concept of granularity in tensor decomposition for the firsttime and propose an HSI recovery model constrained by multi-granularity non-localself-similarity priors. Specifically, the proposed model alternately performscoarse-grained decomposition and fine-grained decomposition on the non-localself-similarity tensor groups. Among them, the coarse-grained decomposition buildsupon Tucker tensor decomposition, which extracts global structural information of theimage by performing singular value shrinkage on the mode-unfolded matrices. Thefine-grained decomposition employs the FCTN decomposition, capturing local detailinformation through modeling pairwise correlations among factor tensors. Thisarchitectural approach achieves a unified representation of global, local, and non-localpriors for HSIs. Experimental results demonstrate that the model has strongapplicability and exhibits outstanding recovery effects in various types of missingscenes such as pixels and stripes.