Abstract
An algorithm for reduction of pulse noises in a scanner (1-dimensional) data is proposed. It is assumed that there is no pulse signal in the data. The algorithm consists of two procedures, that is, noise ‘detection’ and noise ‘reduction’ procedures. The former is based on the fact that the mean value of the data including pulse noises is different from that without them. The difference (therefore the noise) is detected by the statistical t-test on the null hypothesis that the mean values of two sets of data are equal. The criterion for the detection is derived from the statistics for the t-test.
Two kinds of non-linear adaptive noise reduction procedures are proposed here. They are adaptive because they are applied only if the noise is detected. In addition, one of them applies an averaging or a median filter adaptively to the property of the signal, also.
The validity of the algorithm is experimentally confirmed by simulation. The application of the the algorithm to 2-dimensional data is also shown.
