Preliminary elimination of noise from mass spectra by wavelet filters

The paper analyzes applicability of the direct and inverse transformations to the problem of removing noises from mass spectra. To find the low-frequency filter band that ensure optimal results while the resolution remains constant, a multilevel conversion of initial signal is used. As the optimality criterion, an entropy level was taken, which quickly decreases on the early stages of approximation and detailing and decreases much slower on the later stages. The paper presents the cases of filtering an actual mass spectrum by the method suggested.