Abstract
Multispectral images (MSI) contain light information in different wavelengthsof objects, which convey spectral-spatial information and help improve theperformance of various image processing tasks. Numerous techniques have beencreated to extend the application of total variation regularization inrestoring multispectral images, for example, based on channel coupling andadaptive total variation regularization. The primary contribution of this paperis to propose and develop a new multispectral total variation regularization ina generalized opponent transformation domain instead of the originalmultispectral image domain. Here opponent transformations for multispectralimages are generalized from a well-known opponent transformation for colorimages. We will explore the properties of generalized opponent transformationtotal variation (GOTTV) regularization and the corresponding optimizationformula for multispectral image restoration. To evaluate the effectiveness ofthe new GOTTV method, we provide numerical examples that showcase its superiorperformance compared to existing multispectral image total variation methods,using criteria such as MPSNR and MSSIM.