Fig. 2.

GDWavelet Method's Flowchart. Raw MS data is smoothed by bivariate shrinkage estimator in SWT domain to keep true signal and to reduce noise. Without removing baseline, smoothed signal is used to estimate parameters of peaks by zero-crossing lines in multi-scale Gaussian derivative wavelet domain. After removing peaks with frequency response and width less than a threshold, we get all peak candidates. All peak candidates are quantified by PR in envelop analysis and peak height. Union results are final output.