Hyperspectral image denoising by total variation-regularized bilinear factorization

•A novel total variation-regularized bilinear factorization method (BFTV) is proposed for hyperspectral image (HSI) denoising.•BFTV uses the bilinear matrix factorization to replace the original nuclear norm which is a coarse rank approximation.•BFTV utilizes l_1-norm to detect the sparse noise including salt and pepper noise, deadlines, and stripes.•To explore the local piecewise smoothness property in the spatial and spectral domains, BFTV integrates the tensorbased TV regularizer with the bilinear factorization as a unified model.