A least angle regression assessment algorithm based on joint dictionary for visible and near-infrared spectrum denoising

Abstract

Abstract In visible and near-infrared spectroscopic analysis, the noise signals are unavoidable and seriously affect the accuracy and precision of measurement. Considering the characteristics of different components in the mixed spectrum, we propose a least angle regression assessment algorithm based on joint dictionary to remove the noise in this paper. First, the joint dictionary model is constructed to explore the prior knowledge of spectrum, including pure spectrum and noise spectrum characteristics. Then, the least angle regression assessment algorithm is proposed to exploit the difference of the joint dictionary atoms and improve the performance of the model. Finally, two real sample data sets are used to verify the effectiveness of the proposed method. Compared with other methods, the results of the proposed method show outstanding performance in preserving details of the spectral information and significant improvement in denoising.