Robustness of a semi-blind deconvolution method applied to Raman spectra with Gaussian and Poisson noises
In this study, the robustness of semi-blind deconvolution with hybrid regularization (SBD-HR) applied to Raman spectra with Gaussian, Poisson, and Gaussian and Poisson noises is investigated. Gaussian-, Poisson-, and Gaussian and Poisson-degraded cocaine spectra were simulated. SBD-HR was applied to corresponding noise cases of the spectra simulations. Difference spectra, changes in signal-to-noise ratio, values of root-mean-square error, correlation coefficient, weighted correlation coefficient, and standard deviations of distributions fitted to histograms generated from the difference spectra were acquired. Results show that, with the initialized regularization parameters, SBD-HR is capable of resolving Gaussian- and Gaussian and Poisson-degraded spectra with SNR of around 15 and below.