OpenMS  2.7.0
Classes
Spectra Preprocessors

The spectra preprocessors filter the spectra with different criteria. More...

Collaboration diagram for Spectra Preprocessors:

Classes

class  BernNorm
 BernNorm scales the peaks by ranking them and then scaling them according to rank. More...
 
class  MarkerMower
 MarkerMower uses PeakMarker to find peaks, those that are not marked get removed. More...
 
class  NLargest
 NLargest removes all but the n largest peaks. More...
 
class  Normalizer
 Normalizes the peak intensities spectrum-wise. More...
 
class  ParentPeakMower
 ParentPeakMower gets rid of high peaks that could stem from unfragmented precursor ions. More...
 
class  Scaler
 Scaler scales the peak by ranking the peaks and assigning intensity according to rank. More...
 
class  SqrtMower
 Scales the intensity of peaks to the sqrt. More...
 
class  ThresholdMower
 ThresholdMower removes all peaks below a threshold. More...
 
class  WindowMower
 WindowMower augments the highest peaks in a sliding or jumping window. More...
 

Detailed Description

The spectra preprocessors filter the spectra with different criteria.

The preprocessors are able to handle single spectra as well as whole MSExperiments (PeakMap). For single spectra this can be achieved by calling the filterSpectrum or filterPeakSpectrum functions. Additionally PeakMaps can be filtered using the filterPeakMap function. The details of the filters and their options are described in the class documentation respectively.