An Adaptive and Fully Automated Baseline Correction Method for Raman Spectroscopy Based on Morphological Operations and Mollification

Baseline drift is a commonly identified and severe problem in Raman spectra, especially for biological samples. The main cause of baseline drift in Raman spectroscopy is fluorescence generated within the sample. If left untreated, it will affect the following qualitative or quantitative analysis. In this paper, an adaptive and fully automated baseline estimation algorithm based on iteratively averaging morphological opening and closing operations is presented. The proposed method is able to deal with different shapes and amplitudes of baselines. It is tested on both simulated and experimental Raman spectra. Comparison of the proposed method with other morphology-based methods and a well-developed penalized least squares-based method is made. The results demonstrate the superior performance of the proposed method and its advantages—in terms of accuracy, adaptivity, and computing speed—over other algorithms. In general, this method can also be applied to other spectroscopic data or other types of one-dimensional data.