CAJ | 학술논문

针对由光谱仪器测得的拉曼光谱数据经常会受到随机噪声和仪器误差等的影响而导致低分辨率的问题，文中提出了一种能在恢复光谱结构的同时又能抑制光谱噪声的方法，即基于Laplacian-Markov约束的数据加权光谱反卷积模型。该模型将退化光谱中恢复真实光谱的问题转化为最大后验概率的求解问题，推导出了拉曼光谱恢复的变分模型。模型中利用Laplacian-Markov作为光谱数据的光滑性先验，提出加权光谱反卷积来恢复退化的光谱，并使用分裂Bregman迭代法求解。文中对该算法利用实验数据进行了验证，实验表明该方法既能恢复退化光谱细节又能抑制光谱噪声，并且求解速度快，有较强的实用价值。
침대유광보의기측득적랍만광보수거경상회수도수궤조성화의기오차등적영향이도치저분변솔적문제，문중제출료일충능재회복광보결구적동시우능억제광보조성적방법，즉기우Laplacian-Markov약속적수거가권광보반권적모형。해모형장퇴화광보중회복진실광보적문제전화위최대후험개솔적구해문제，추도출료랍만광보회복적변분모형。모형중이용Laplacian-Markov작위광보수거적광활성선험，제출가권광보반권적래회복퇴화적광보，병사용분렬Bregman질대법구해。문중대해산법이용실험수거진행료험증，실험표명해방법기능회복퇴화광보세절우능억제광보조성，병차구해속도쾌，유교강적실용개치。
Raman spectroscopic data often suffers from common problems of bands overlapping and random Gaussian noise. Spectral resolution can be improved by mathematically removing the effect of the instrument response function. In this paper, a novel method to deconvolute the degraded spectrum with the Laplacian-Markov priori was proposed, solving by split Bregman optimization scheme, which was fast, robust to noise and easy to implement. The Laplacian-Markov priori was proposed to save the shape peaks and suppress the noise. A data weighted operator was introduced to spectral deconvolution to find a balance between band narrowing and noise suppression. The method could estimate spectral structural details as well as suppress the noise effectively. Experimental results with real Raman spectra manifest that this algorithm can deconvolute the overlapping peaks as well as suppress the noise effectively. Owing to the fast of computing time, it is expected that the proposed method has considerable value in practice.