Hyperspectral images (HSI) are contaminated with a large amount of mixed noise during acquisition due to the environmental and sensor influences,which can affect the performance of the subsequent application of remote sensing image.Therefore,recovering clean HSI from mixed noise has become an important pre processing process.However,some existing tensor models cannot achieve good results in removing the mixed noise containing bands and deadlines.To this end,a hyperspectral mixed noise image restoration algorithm (FRTCSSTV) based on group sparse spatial spectral total variation is proposed.In order to avoid excessive smoothing,group sparse spatial spectral total variation regularization is used to enhance the sparsity of spatial spectral dimension.At the same time,to maintain the original structure of HSI,the global low rank of HSI is expressed by directly constraining the rank of tensor fibers.In both simulated and real hyperspectral image experiments,the FRTCSSTV method has better performance in removing mixed noise containing stripe and dead line noise compared to other models.