Multiplex Transformed Tensor Decomposition Based Single Hyperspectral Image Super-resolution for IgA Diagnostic Applications

Hyperspectral imaging brings a new pattern for the diagnosis and pathological evaluation of IgA nephropathy. However, the low spatial resolution of hyperspectral becomes an obstacle to accurate labeling and diagnosis. Considering the characteristics of medical hyperspectral data, we propose a tensor based super-resolution reconstruction method. To makes full use of the correlations along all modes in IgA hyperspectral images, the MTTD framework is applied in the model. 3-D TV is also utilized to constrain structural smoothness. The experimental results show that the reconstruction results of the model are valid. The tensor-based super-resolution method provides a new preprocessing method for the pathological study of IgA nephropathy combined with hyperspectral imaging, and has potential clinical value for the intelligent analysis of high-dimensional medical data.