| Abstract
| - Multiplexing and time averaging of signal are effectivenoise reduction protocols applied in many analyticalmeasurement systems. The efficacy of these protocols maybe reduced by random occurrences of high-magnitudenoise that do not conform to the statistical distribution ofnoise for all other measurements in the data set. Thishigh-magnitude noise, which may have an insignificantprobability of occurrence for a single measurement,almost certainly affects data collected in a multichannel,multiplexed modality, such as Fourier transform infrared(FT-IR) spectroscopic imaging employing focal plane arraydetectors. To recover time-averaging advantages in thesecases, we present a general coaddition method that usestwo statistical measures, the mean and median of theensemble of measurements of a signal, to obtain a betterestimate of the true signal than that estimated by timeaveraging alone. This method, termed median filtered timeaveraging, is shown to be an effective noise removalprocedure for FT-IR imaging data. The effects of noiseremoval on time averaging and multiplexing are examinedtheoretically and are demonstrated for hyperspectralinfrared microspectroscopic imaging data obtained fromhuman skin biopsies by using a rapid data acquisitionprocedure.
|
