Regularized Tensor Representative Coefficient Model for Hyperspectral Target Detection

Target detection based on hyperspectral image (HSI) representations has drawn wide attention given its variety of features. The matrix-based approach inevitably loses spatial information and fails to explore the intrinsic multimodal structure an HSI cube. In this paper, we propose a regularized tensor-based model without altering data structure. We assume that observed third-order tensor is decomposed into sum Total Variation Low-rank background Sparse (TVLrS) target tensor. two tensors are represented as mode-3 product tensor, called Tensor Representation Coefficient (TRC), spectra dictionary matrix. Then, coined TVLrS-TRC. TRC low-rank property, contributing low-rankness characterization in our model. Moreover, size term smaller than characterizing local smoothness via TV regularization reduces computational cost compared Extensive experiments real datasets demonstrate advantage proposed method with state-of-the-art.