Color image denoising via monogenic matrix-based sparse representation

Abstract

Traditional sparse representation models usually treat color image channels independently, which ignore the relationship among the channels, resulting in inaccurate sparse coding coefficients. In this paper, we propose a novel vector-valued sparse representation model for color images using monogenic matrix. The proposed model treats color image as a monogenic matrix, which can transform independent color channels into a whole. In the dictionary learning stage, a corresponding effective dictionary learning method is designed by using monogenic matrix singular value decomposition, which conducts sparse basis selection in monogenic space. Then, a monogenic-based orthogonal matching pursuit algorithm is presented in the sparse coding stage. In order to demonstrate the effectiveness of the proposed sparse representation model, we apply the model to color image denoising. Extensive experimental results on color image denoising manifest that the proposed model outperforms the state-of-the-art schemes.