Automatic data editing for high-amplitude noise

Chapter  addresses the removal of high-amplitude noise. While this topic is somewhat different from the rest of the thesis, data editing is often a step needed before the other processes can be attempted. Removing these high-amplitude samples is done by comparing a trace with its neighbors using small two-dimensional prediction-error filters. Each prediction is done against a single nearby trace. In the two-dimensional problem, two predictions are done, one for each of a trace's nearest neighbors. The best predictions are taken from these two predictions, and if the best prediction exceeds some value, the corresponding input sample is removed and marked as having been deleted. These muted samples may be later restored as discussed in chapter .